Systems and methods for detection of structures and/or patterns in images
The subject disclosure presents systems and computer-implemented methods for automatic immune cell detection that is of assistance in clinical immune profile studies. The automatic immune cell detection method involves retrieving a plurality of image channels from a multi-channel image such as an RGB image or biologically meaningful unmixed image. A cell detector is trained to identify the immune cells by a convolutional neural network in one or multiple image channels. Further, the automatic immune cell detection algorithm involves utilizing a non-maximum suppression algorithm to obtain the immune cell coordinates from a probability map of immune cell presence possibility generated from the convolutional neural network classifier.
1. A computer-implemented method comprising:
obtaining a training image of a biological specimen;
separating, using color deconvolution, the training image into a plurality of color channels;
obtaining a label for a biological structure in the biological specimen, wherein the label provides an identity and locator of the cellular structure, and the locator provides the location of the biological structure within the training image;
extracting an image patch around the location of the biological structure from each color channel of the plurality of color channels;
identifying the image patch and the label as a ground truth; and
training a convolutional neural network model based on the ground truth.
2. The method of claim 1 , wherein the locator includes coordinates or selected pixel or pixels of a center or centroid of the biological structure.
3. The method of claim 2 , wherein the locator is centered or substantially centered within the image patch, and wherein a size of the image patch corresponds to a size of the biological structure relative to a size of the training image.
4. The method of claim 3 , further comprising generating a trained convolutional neural network model for the biological structure based on the training of the convolutional neural network with the ground truth.
5. The method of claim 4 , wherein the plurality of color channels comprise at least a cellular structure channel and a background image structure channel.
6. The method of claim 5 , wherein the image patch is extracted around the location of the biological structure determined by radial symmetry, ring detection, or foreground thresholding.
7. The method of claim 1 , wherein the biological specimen is stained with a plurality of stains and each color channel of the plurality of color channels corresponds to a different stain from the plurality of stains used to stain the biological specimen.
8. A system comprising:
one or more data processors; and
a non-transitory computer readable storage medium containing instructions which, when executed on the one or more data processors, cause the one or more data processors to perform actions including:
obtaining a training image of a biological specimen;
separating, using color deconvolution, the training image into a plurality of color channels;
obtaining a label for a biological structure in the biological specimen, wherein the label provides an identity and locator of the cellular structure, and the locator provides the location of the biological structure within the training image;
extracting an image patch around the location of the biological structure from each color channel of the plurality of color channels;
identifying the image patch and the label as a ground truth; and
training a convolutional neural network model based on the ground truth.
9. The system of claim 8 , wherein the locator includes coordinates or selected pixel or pixels of a center or centroid of the biological structure.
10. The system of claim 9 , wherein the locator is centered or substantially centered within the image patch, and wherein a size of the image patch corresponds to a size of the biological structure relative to a size of the training image.
11. The system of claim 10 , further comprising generating a trained convolutional neural network model for the biological structure based on the training of the convolutional neural network with the ground truth.
12. The system of claim 11 , wherein the plurality of color channels comprise at least a cellular structure channel and a background image structure channel.
13. The system of claim 12 , wherein the image patch is extracted around the location of the biological structure determined by radial symmetry, ring detection, or foreground thresholding.
14. The system of claim 8 , wherein the biological specimen is stained with a plurality of stains and each color channel of the plurality of color channels corresponds to a different stain from the plurality of stains used to stain the biological specimen.
15. A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause one or more data processors to perform actions including:
obtaining a training image of a biological specimen;
separating, using color deconvolution, the training image into a plurality of color channels;
obtaining a label for a biological structure in the biological specimen, wherein the label provides an identity and locator of the cellular structure, and the locator provides the location of the biological structure within the training image;
extracting an image patch around the location of the biological structure from each color channel of the plurality of color channels;
identifying the image patch and the label as a ground truth; and
training a convolutional neural network model based on the ground truth.
16. The computer-program product of claim 15 , wherein the locator includes coordinates or selected pixel or pixels of a center or centroid of the biological structure.
17. The computer-program product of claim 16 , wherein the locator is centered or substantially centered within the image patch, and wherein a size of the image patch corresponds to a size of the biological structure relative to a size of the training image.
18. The computer-program product of claim 17 , further comprising generating a trained convolutional neural network model for the biological structure based on the training of the convolutional neural network with the ground truth.
19. The computer-program product of claim 18 , wherein the biological specimen is stained with a plurality of stains and each color channel of the plurality of color channels corresponds to a different stain from the plurality of stains used to stain the biological specimen.
20. The computer-program product of claim 15 , wherein the image patch is extracted around the location of the biological structure determined by radial symmetry, ring detection, or foreground thresholding.