Typing biological cells
A system for typing biological cells includes a tunable Fabry-Perot etalon, and imaging sensor, and a processor. The imaging sensor acquires one or more images of one or more biological cells from light transmitted through the tunable Fabry-Perot etalon. Each image represents signal associated with one or more wavelengths transmitted through the tunable Fabry-Perot etalon. The processor is configured to determine a type of each of the one or more biological cells. Determining the type uses a machine learning algorithm and is based at least in part on one or more of an image segmentation, a patch extraction, a feature extraction, a feature compression, a deep feature extraction, a feature fusion, a feature classification, and a prediction map reconstruction.
1. A system, comprising:
a tunable Fabry-Perot etalon;
an imaging sensor, wherein the imaging sensor acquires one or more images of one or more biological cells from light transmitted through the one or more biological cells and separately through the tunable Fabry-Perot etalon, wherein each image represents a signal associated with one or more wavelengths transmitted through the tunable Fabry-Perot etalon; and
a processor configured to:
determine a type of each of the one or more biological cells, wherein determining the type uses a machine learning algorithm, and wherein determining the type is based at least in part on one or more of an image segmentation, a patch extraction, a feature extraction, a feature compression, a deep feature extraction, a feature fusion, a feature classification, and a prediction map reconstruction.
2. The system as in claim 1 , wherein the image segmentation comprises delineation of pixels belonging to the one or more biological cells in the one or more images.
3. The system as in claim 2 , wherein the delineation of the pixels comprises determining a foreground and a background, and wherein the foreground and the background are used to generate a binary segmentation mask.
4. The system as in claim 3 , wherein morphology and contours in the binary segmentation mask are used in determining the patch extraction.
5. The system as in claim 1 , wherein the patch extraction comprises cropping the one or more images to generate one or more patch images comprising an individual biological cell image and/or a cluster of biological cells image.
6. The system as in claim 5 , wherein the patch extraction generates a cell-cluster mask for each individual biological cell image and each cluster of biological cells image.
7. The system as in claim 6 , wherein the feature extraction comprises determining a first set of one or more morphological features of interest from the cell-cluster mask to generate a first list of feature metrics.
8. The system as in claim 7 , wherein the feature compression comprises spectrally compressing the cell-cluster mask to generate a compressed cell-cluster mask.
9. The system as in claim 8 , wherein the deep feature extraction comprises determining a second set of one or more morphological features of interest from the compressed cell-cluster mask to generate a second list of feature metrics.
10. The system as in claim 9 , wherein the feature fusion comprises appending the second list of feature metrics to the first list of feature metrics to generate a complete list of feature metrics.
11. The system as in claim 10 , wherein the feature classification comprises determining the type of biological cell in the cell-cluster mask based at least in part on comparing the complete list of feature metrics to a library of feature metrics corresponding to known biological cell types.
12. The system as in claim 11 , wherein the prediction map reconstruction comprises associating the type of biological cell in the cell-cluster mask with its location within the one or more images of one or more biological cells.
13. The system as in claim 12 , wherein the prediction map reconstruction is used to improve confidence in the type of biological cell determined from the cell-cluster mask by comparing to one or more cell-cluster masks of the individual biological cell image and/or the cluster of biological cells image.
14. The system as in claim 13 , wherein determining the type utilizes biological cell shape and/or size information as determined from the cell-cluster mask.
15. The system as in claim 1 , wherein the type of each of the one or more biological cells comprises one or more of the following: a eukaryote, a prokaryote, a pathogen, a bacterium, an archaeon, a fungus, a plant, an animal, a human, a protist, a slime mold, a protozoon, an algae, a yeast, a species, a sub-species, a serotype, or a strain.
16. The system as in claim 1 , wherein the one or more images are acquired using a microscope.
17. The system as in claim 1 , wherein the image sensor comprises a solid-state sensor, a CMOS sensor, a CCD sensor, a staring array, an RGB sensor, an IR sensor, a Bayer pattern color sensor, a multiple band sensor, or a monochrome sensor.
18. The system as in claim 1 , wherein the processor utilizes one or more of the following machine learning algorithms to determine the type: a neural network model, a bounding box model, a clustering algorithm, or a classifier algorithm.
19. A method, comprising:
receiving one or more images of one or more biological cells acquired using an image sensor from light transmitted through the one or more biological cells and separately through a tunable Fabry-Perot etalon, wherein each image represents a signal associated with one or more wavelengths transmitted through the tunable Fabry-Perot etalon; and
determining, using a processor, a type of each of the one or more biological cells, wherein determining the type uses a machine learning algorithm, and wherein determining the type is based at least in part on one or more of an image segmentation, a patch extraction, a feature extraction, a feature compression, a deep feature extraction, a feature fusion, a feature classification, and a prediction map reconstruction.
20. A non-transitory computer readable storage medium comprising computer instructions for:
receiving one or more images of one or more biological cells acquired using an image sensor from light transmitted through the one or more biological cells and separately through a tunable Fabry-Perot etalon, wherein each image represents a signal associated with one or more wavelengths transmitted through the tunable Fabry-Perot etalon; and
determining, using a processor, a type of each of the one or more biological cells, wherein determining the type uses a machine learning algorithm, and wherein determining the type is based at least in part on one or more of an image segmentation, a patch extraction, a feature extraction, a feature compression, a deep feature extraction, a feature fusion, a feature classification, and a prediction map reconstruction.