Image classifier training
View Patent ↗Methods are disclosed that include: (a) applying a first stain to a first sample having a plurality of regions, where the first stain selectively binds to only a first subset of the regions of the first sample; (b) applying a second stain to the first sample, where the second stain binds to a second set of regions of the first sample; (c) obtaining an image of the first sample, and analyzing the image to obtain a first component image corresponding substantially only to spectral contributions from the first stain, and a second component image corresponding substantially only to spectral contributions from the second stain; and (d) training a classifier to identify regions of a second sample based on information derived from the first and second component images, the identified regions corresponding to the first subset of regions of the first sample.
1. A system, comprising:
a support apparatus configured to support samples;
a radiation source configured to illuminate samples;
an imaging apparatus configured to obtain images of samples; and
an electronic processor connected to the imaging apparatus and configured to:
receive an image of a first sample on the support apparatus, the first sample comprising a first stain and a second stain, wherein the first stain selectively binds to only a first subset of regions of the first sample;
analyze the image of the first sample to obtain a first component image corresponding substantially only to spectral contributions from the first stain, and a second component image corresponding substantially only to spectral contributions from the second stain; and
train a classifier to identify a second subset of regions in a second sample that corresponds to the first subset of regions in the first sample, wherein the second sample does not comprise the first stain.
2. The system of claim 1 , wherein the electronic processor is further configured to:
receive an image of the second sample on the support apparatus;
analyze the image of the second sample to obtain a third component image corresponding substantially only to spectral contributions from the second stain in the second sample; and
use the classifier to identify the second subset of regions in the second sample based on the third component image.
3. The system of claim 2 , wherein the second stain binds unselectively to the second sample, and wherein the image of the second sample comprises contributions from the second stain in substantially all regions of the second sample.
4. The system of claim 1 , wherein the first and second subsets of regions correspond to a common type of cell or sub-cellular compartment.
5. The system of claim 1 , wherein the first and second subsets of regions correspond to a common type of tissue structure.
6. The system of claim 1 , wherein the first and second subsets of regions correspond to regions where a common biological or chemical agent is localized.
7. The system of claim 1 , wherein the first and second subsets of regions correspond to membranes in the first and second samples.
8. The system of claim 1 , wherein the first and second subsets of regions correspond to cell nuclei in the first and second samples.
9. The system of claim 1 , wherein the first stain comprises at least one member of the group consisting of immunohistochemical agents, cytokeratin, and cadherin.
10. The system of claim 1 , wherein the second stain comprises at least one member of the group consisting of immunohistochemical agents, DAPI, Hoechst blue, hematoxylin, DAB, fast red, fluorescein, rhodamine, Texas red, Cy3, Cy5, Cy5.5, Alexa dyes, and quantum dots.
11. The system of claim 1 , wherein the first sample comprises a third stain, and wherein the electronic processor is further configured to analyze the image of the first sample to obtain a third component image corresponding substantially only to spectral contributions from the third stain.
12. The system of claim 11 , wherein:
the third stain selectively binds to only a third subset of regions of the first sample;
the electronic processor is further configured to train the classifier to identify a fourth subset of regions in the second sample that corresponds to the third subset of regions in the first sample; and
the second sample does not comprise the third stain.
13. The system of claim 12 , wherein the first and third subsets of regions of the first sample have at least some portions of the first sample in common.
14. The system of claim 1 , wherein the electronic processor is configured to analyze the image of the first sample by unmixing the image to obtain the first and second component images.
15. The system of claim 1 , wherein the electronic processor is configured to analyze the image of the first sample by separating color channels of the image to obtain the first and second component images.
16. The system of claim 1 , wherein the second stain binds unselectively to the first sample, and wherein the second component image comprises contributions from the second stain in substantially all regions of the first sample.
17. The system of claim 1 , wherein the imaging apparatus is configured to obtain images of samples by detecting fluorescence emission from the samples, incident radiation reflected by the samples, or incident radiation transmitted by the samples.
18. The system of claim 1 , wherein the electronic processor is configured to train the classifier by providing position information about the first subset of regions to the classifier.
19. The system of claim 18 , wherein the position information comprises texture information derived from the image of the first sample.
20. The system of claim 1 , wherein the classifier comprises a machine learning classifier.
21. The system of claim 1 , wherein the electronic processor is further configured to:
analyze the image of the first sample to obtain a component image corresponding to autofluorescence of the first sample; and
train the classifier to identify the second subset of regions in the second sample that corresponds to the first subset of regions of the first sample based on autofluorescence information about the second sample.
22. A system, comprising:
a support apparatus configured to support samples;
a radiation source configured to illuminate samples;
an imaging apparatus configured to obtain images of samples; and
an electronic processor connected to the imaging apparatus and configured to:
receive an image of a first sample on the support apparatus, the first sample comprising n selective stains and m non-selective stains, wherein each of the n selective stains binds to only a subset of regions of the first sample;
analyze the image of the first sample to obtain (n+m) component images, each of the component images corresponding substantially only to spectral contributions from one of the (n+m) stains; and
for at least one of the n selective stains, train a classifier to identify a subset of regions in a second sample that corresponds to a subset of regions of the first sample to which the at least one selective stain binds, wherein the second sample does not comprise any of the n selective stains,
wherein m≧1 and n≧m.
23. The system of claim 22 , wherein the electronic processor is further configured to:
receive an image of the second sample on the support apparatus;
analyze the image of the second sample to obtain a component image corresponding substantially only to spectral contributions from one of the m non-selective stains in the second sample; and
use the classifier to identify the subset of regions in the second sample based on the component image corresponding to the second sample.
24. The system of claim 22 , wherein the subsets of regions in the first and second samples correspond to a common type of cell or sub-cellular compartment.
25. The system of claim 22 , wherein the subsets of regions in the first and second samples correspond to a common type of tissue structure.
26. The system of claim 22 , wherein the subsets of regions in the first and second samples correspond to regions where a common biological or chemical agent is localized.
27. The system of claim 22 , wherein each of the n selective stains comprises a member of the group consisting of immunohistochemical agents, cytokeratin, and cadherin.
28. The system of claim 22 , wherein each of the m non-selective stains comprises a member of the group consisting of immunohistochemical agents, DAPI, Hoechst blue, hematoxylin, DAB, fast red, fluorescein, rhodamine, Texas red, Cy3, Cy5, Cy5.5, Alexa dyes, and quantum dots.
29. The system of claim 22 , wherein m=1.
30. The system of claim 29 , wherein the first and second samples each comprise the n selective stains, and wherein n≧2.
31. The system of claim 22 , wherein m=2.
32. The system of claim 31 , wherein the first and second samples each comprise the n selective stains, and wherein n≧2.
33. The system of claim 22 , wherein the second sample comprises each of the n selective stains, and for each one of the n selective stains, the electronic processor is configured to train the classifier to identify a subset of regions in the second sample that corresponds to a subset of regions of the first sample to which the each one of the n selective stains selectively binds.
34. The system of claim 22 , wherein the electronic processor is configured to analyze the image of the first sample by unmixing the image to obtain the (n+m) component images.
35. The system of claim 22 , wherein the electronic processor is configured to analyze the image of the first sample by separating color channels of the image to obtain the (n+m) component images.
36. The system of claim 22 , wherein the electronic processor is further configured to:
analyze the image of the first sample to obtain a component image corresponding to autofluorescence of the first sample; and
for the at least one of the n selective stains, train the classifier to identify the subset of regions in the second sample that corresponds to the subset of regions of the first sample to which the at least one selective stain binds based on autofluorescence information about the second sample.