Identifying microorganisms using three-dimensional quantitative phase imaging

1 . A method comprising:

using a phase-contrast microscope to generate a three-dimensional (3D) quantitative phase image (QPI) of one or more microorganisms, wherein the 3D QPI comprises a 3D representation of the microorganisms; and

generating a latent feature representation of the microorganisms using a neural network, comprising:

processing the 3D QPI using the neural network to generate the latent feature representation of the microorganisms as an intermediate output generated by one or more intermediate layers of the neural network, wherein each intermediate layer is a hidden layer of the neural network, wherein:

the neural network is a convolutional neural network comprising one or more 3D convolutional layers; and

the neural network has been trained to process an input 3D QPI in accordance with current values of a set of neural network parameters of the neural network to generate a neural network output characterizing the microorganisms;

the neural network has been trained perform a plurality of different prediction tasks, including two or more of: predicting a species of microorganisms in the input 3D QPI, predicting a strain of microorganisms in the input 3D QPI, predicting a gram-stainability of microorganisms in the input 3D QPI, predicting a metabolism of microorganisms in the input 3D QPI, predicting a morphology of microorganisms in the input 3D QPI, or predicting a motility of microorganisms in the input 3D QPI; and

processing the latent feature representation of the microorganisms using a separate prediction machine learning model that is different from the neural network to generate a prediction characterizing the microorganisms.

2 . The method of claim 1 , wherein:

the neural network output comprises a respective probability value for each of a predetermined number of microorganism types; and

the probability value for a given microorganism type indicates a likelihood that the microorganisms are of the given microorganism type.

3 . The method of claim 1 , wherein the neural network output comprises a probability value of a microorganism type that indicates a likelihood that the microorganisms are of the microorganism type.

4 . The method of claim 1 , wherein the predicted type of the microorganisms is selected from the group consisting of genus, species, strain, gram-stainability, metabolism, morphology, and motility.

5 . The method of claim 1 , wherein the three-dimensional quantitative phase image of the microorganisms is a three-dimensional refractive index tomogram.

6 . The method of claim 1 , wherein the microorganisms are selected from the group consisting of bacteria, viruses, fungi, parasites, and microalgae.

7 . The method of claim 6 , wherein the microorganisms comprise bacteria.

8 . The method of claim 7 , wherein the bacteria are present in a blood sample of a patient.

9 . The method of claim 8 , further comprising administering an antibiotic therapy to the patient based on the predicted type of the bacteria.

10 . The method of claim 1 , wherein using a phase-contrast microscope to generate a three-dimensional quantitative phase image of the microorganisms comprises:

using the phase-contrast microscope to generate phase and amplitude images of the microorganisms at each of a plurality of illumination angles; and

reconstructing a three-dimensional refractive index tomogram using the phase and amplitude images.

11 . The method of claim 1 , wherein the method takes at most one hour.

12 . The method of claim 1 , wherein:

the one or more microorganisms are isolated from a biological sample from a patient with a bacterial infection or that is suspected of having a bacterial infection; and

the prediction characterizing the microorganisms is generated within one hour of obtaining the biological sample from the patient.

13 . The method of claim 12 , wherein the biological sample comprises a blood sample.

14 . The method of claim 12 , wherein the prediction characterizing the microorganisms is generated within 45 minutes of obtaining the biological sample from the patient.

15 . The method of claim 12 , wherein the prediction characterizing the microorganisms is generated within 30 minutes of obtaining the biological sample from the patient.

16 . The method of claim 12 , wherein the prediction characterizing the microorganisms is generated within 15 minutes of providing the biological sample from the patient.

17 . One or more non-transitory computer readable storage media storing instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:

receiving a three-dimensional (3D) quantitative phase image (QPI) of one or more microorganisms that is generated using a phase-contrast microscope, wherein the 3D QPI comprises a 3D representation of the microorganisms; and

generating a latent feature representation of the microorganisms using a neural network, comprising:

processing the 3D QPI using the neural network to generate the latent feature representation of the microorganisms as an intermediate output generated by one or more intermediate layers of the neural network, wherein each intermediate layer is a hidden layer of the neural network, wherein:

the neural network is a convolutional neural network comprising one or more 3D convolutional layers; and

the neural network has been trained to process an input 3D QPI in accordance with current values of a set of neural network parameters of the neural network to generate a neural network output characterizing the microorganisms;

the neural network has been trained perform a plurality of different prediction tasks, including two or more of: predicting a species of microorganisms in the input 3D QPI, predicting a strain of microorganisms in the input 3D QPI, predicting a gram-stainability of microorganisms in the input 3D QPI, predicting a metabolism of microorganisms in the input 3D QPI, predicting a morphology of microorganisms in the input 3D QPI, or predicting a motility of microorganisms in the input 3D QPI; and

processing the latent feature representation of the microorganisms using a separate prediction machine learning model that is different from the neural network to generate a prediction characterizing the microorganisms.

18 . A system comprising:

one or more computers; and

one or more storage devices communicatively coupled to the one or more computers, wherein the one or more storage devices store instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising:

receiving a three-dimensional (3D) quantitative phase image (QPI) of one or more microorganisms that is generated using a phase-contrast microscope, wherein the 3D QPI comprises a 3D representation of the microorganisms; and

generating a latent feature representation of the microorganisms using a neural network, comprising:

processing the 3D QPI using the neural network to generate the latent feature representation of the microorganisms as an intermediate output generated by one or more intermediate layers of the neural network, wherein each intermediate layer is a hidden layer of the neural network, wherein:

the neural network is a convolutional neural network comprising one or more 3D convolutional layers; and

the neural network has been trained to process an input 3D QPI in accordance with current values of a set of neural network parameters of the neural network to generate a neural network output characterizing the microorganisms;

the neural network has been trained perform a plurality of different prediction tasks, including two or more of: predicting a species of microorganisms in the input 3D QPI, predicting a strain of microorganisms in the input 3D QPI, predicting a gram-stainability of microorganisms in the input 3D QPI, predicting a metabolism of microorganisms in the input 3D QPI, predicting a morphology of microorganisms in the input 3D QPI, or predicting a motility of microorganisms in the input 3D QPI; and

processing the latent feature representation of the microorganisms using a separate prediction machine learning model that is different from the neural network to generate a prediction characterizing the microorganisms.

19 . The method of claim 1 , wherein the prediction machine learning model is a non-differentiable machine learning model.

20 . The method of claim 1 , wherein the prediction machine learning model comprises a random forest model.

21 . The method of claim 1 , wherein the prediction machine learning model comprises a support vector machine model.

22 . The method of claim 1 , wherein the one or more intermediate layers of the neural network that generate the latent feature representation of the microorganisms comprise at least one fully-connected intermediate layer of the neural network.

23 . The system of claim 18 , wherein:

the neural network output comprises a respective probability value for each of a predetermined number of microorganism types; and

the probability value for a given microorganism type indicates a likelihood that the microorganisms are of the given microorganism type.