Stain-free histopathology by chemical imaging
The present disclosure provides methods, systems, and computer-readable storage media that can be used to image an unstained sample. The disclosed methods can include obtaining a spectroscopic image (e.g., infrared (IR) imaging data) of the sample, analyzing the resulting spectroscopic image to reduce the dimensionality of the spectroscopic image, comparing the reduced spectroscopic image compared to a control (e.g., by using an appropriately trained algorithm) and generating an output computed stain image from the reduced IR spectra, thereby imaging the sample without the use of stains or dyes.
1. A method of imaging a single unstained sample, comprising:
identifying a group of control samples corresponding to a target stain, wherein the control samples include spectroscopic data and images of a plurality of samples that are stained with the target stain;
generating a neural network from the group of control samples, wherein the neural network is a feed-forward neural network trained using reduced spectroscopic data from the group of control samples as input, wherein the neural network generates bright-field color data as output;
obtaining a spectroscopic image of the single unstained sample;
reducing dimensionality of the spectroscopic image, thereby generating reduced spectroscopic image data; and
generating an output computed stain image from the reduced spectroscopic image data using the neural network, thereby imaging the unstained sample to represent of the target stain.
2. The method of claim 1 , wherein the spectroscopic image of the single unstained sample is obtained using IR spectroscopic imaging instrumentation.
3. The method of claim 1 , wherein the spectroscopic images of the single unstained sample are obtained using a Fourier transform infrared spectrometer.
4. The method of claim 1 , wherein obtaining the spectroscopic images of the single unstained sample comprises obtaining two spatial dimensions and one spectral dimension.
5. The method of claim 1 , wherein reducing dimensionality of the spectroscopic image comprises use of principal component analysis (PCA) or a metrics approach.
6. The method of claim 1 , further comprising obtaining a bright-field image of the single unstained sample.
7. The method of claim 1 , wherein the method detects one or more target proteins or target pathogens.
8. The method of claim 1 , wherein the output computed stain image is comparable to an image obtained from staining the single unstained sample.
9. The method of claim 1 , wherein the single unstained sample is a biological sample from a mammal.
10. The method of claim 9 , wherein the single unstained sample is known or suspected of comprising cancer.
11. The method of claim 10 , wherein the cancer is breast, lung, renal, pancreas, prostate, colon, rectal, ovary, or liver cancer.
12. The method of claim 1 , wherein the single unstained sample is from an environmental or food source.
13. The method of claim 9 , wherein the single unstained sample is known or suspected of comprising a pathogen.
14. The method of claim 13 , wherein the pathogen is a bacterium, virus, fungi, protozoa, or bacterial spore.
15. The method of claim 1 , wherein the single unstained sample is a fixed, fresh or frozen tissue sample.
16. The method of claim 10 , further comprising treating a subject identified as having cancer.
17. The method of claim 1 , further comprising selecting a subject having or suspected of having cancer and obtaining a tissue sample from the subject.
18. A non-transitory computer-readable storage medium having instructions thereon that, when executed by a processing system including a processor, facilitate performance of operations, comprising:
obtaining a spectroscopic image of an unstained sample;
reducing dimensionality of the spectroscopic image by ignoring particular data of the spectroscopic image, thereby generating a reduced spectroscopic image, wherein the particular data being ignored is determined according to pixels that do not satisfy a threshold absorption;
inputting the reduced spectroscopic image into a neural network;
comparing the reduced spectroscopic image to parameters from the neural network of a control sample, wherein the control sample corresponds to a staining process; and
generating an output computed stain image from the reduced spectroscopic image based on the comparing.
19. A device, comprising:
a processing system including a processor; and
a memory that stores executable instructions that, when executed by the processing system, facilitate performance of operations, comprising:
training a feed-forward network using spectroscopic images from a plurality of training samples as input and associated bright-field color data of a staining process applied to the training samples as output parameters;
obtaining a spectroscopic image of an unstained sample;
reducing dimensionality of the spectroscopic image by ignoring particular data of the spectroscopic image, thereby generating a reduced spectroscopic image;
comparing the reduced spectroscopic image to output parameters of a control sample, wherein the control sample corresponds to a particular staining process; and
generating an output computed stain image from the reduced spectroscopic image based on the comparing.