Phenotype analysis of cellular image data using a deep metric network
First Claim
1. A method, comprising:
- receiving, by a computing device, a target image of a target biological cell having a target phenotype;
obtaining, by the computing device, a semantic embedding associated with the target image, wherein the semantic embedding associated with the target image is generated using a machine-learned, deep metric network model;
obtaining, by the computing device for each of a plurality of candidate images of candidate biological cells each having a respective candidate phenotype, a semantic embedding associated with the respective candidate image, wherein the semantic embedding associated with the respective candidate image is generated using the machine-learned, deep metric network model;
determining, by the computing device, a similarity score for each candidate image,wherein determining the similarity score for a respective candidate image comprises computing, by the computing device, a vector distance in a multi-dimensional space described by the semantic embeddings between the respective candidate image and the target image, andwherein the similarity score for each candidate image represents a degree of similarity between the target phenotype and the respective candidate phenotype;
determining, by the computing device, a threshold similarity score; and
determining, by the computing device, those candidate images having similarity scores that satisfy the threshold similarity score,wherein the target phenotype is a healthy phenotype,wherein the candidate images having similarity scores that satisfy the threshold similarity score have respective candidate phenotypes corresponding to the healthy phenotype, andwherein those candidate images having similarity scores that do not satisfy the threshold similarity score have respective candidate phenotypes corresponding to an unhealthy phenotype.
2 Assignments
0 Petitions
Accused Products
Abstract
The disclosure relates to phenotype analysis of cellular image data using a machine-learned, deep metric network model. An example method includes receiving, by a computing device, a target image of a target biological cell having a target phenotype. Further, the method includes obtaining, by the computing device, semantic embeddings associated with the target image and each of a plurality of candidate images of candidate biological cells each having a respective candidate phenotype. The semantic embeddings are generated using a machine-learned, deep metric network model. In addition, the method includes determining, by the computing device, a similarity score for each candidate image. Determining the similarity score for a respective candidate image includes computing a vector distance between the respective candidate image and the target image. The similarity score for each candidate image represents a degree of similarity between the target phenotype and the respective candidate phenotype.
-
Citations
24 Claims
-
1. A method, comprising:
-
receiving, by a computing device, a target image of a target biological cell having a target phenotype; obtaining, by the computing device, a semantic embedding associated with the target image, wherein the semantic embedding associated with the target image is generated using a machine-learned, deep metric network model; obtaining, by the computing device for each of a plurality of candidate images of candidate biological cells each having a respective candidate phenotype, a semantic embedding associated with the respective candidate image, wherein the semantic embedding associated with the respective candidate image is generated using the machine-learned, deep metric network model; determining, by the computing device, a similarity score for each candidate image, wherein determining the similarity score for a respective candidate image comprises computing, by the computing device, a vector distance in a multi-dimensional space described by the semantic embeddings between the respective candidate image and the target image, and wherein the similarity score for each candidate image represents a degree of similarity between the target phenotype and the respective candidate phenotype; determining, by the computing device, a threshold similarity score; and determining, by the computing device, those candidate images having similarity scores that satisfy the threshold similarity score, wherein the target phenotype is a healthy phenotype, wherein the candidate images having similarity scores that satisfy the threshold similarity score have respective candidate phenotypes corresponding to the healthy phenotype, and wherein those candidate images having similarity scores that do not satisfy the threshold similarity score have respective candidate phenotypes corresponding to an unhealthy phenotype. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22)
-
-
23. A non-transitory, computer-readable medium having instructions stored thereon, wherein the instructions, when executed by a processor, cause the processor to execute a method, comprising:
-
receiving, by the processor, a target image of a target biological cell having a target phenotype; obtaining, by the processor, a semantic embedding associated with the target image, wherein the semantic embedding associated with the target image is generated using a machine-learned, deep metric network model; obtaining, by the processor for each of a plurality of candidate images of candidate biological cells each having a respective candidate phenotype, a semantic embedding associated with the respective candidate image, wherein the semantic embedding associated with the respective candidate image is generated using the machine-learned, deep metric network model; determining, by the processor, a similarity score for each candidate image, wherein determining the similarity score for a respective candidate image comprises computing, by the processor, a vector distance in a multi-dimensional space described by the semantic embeddings between the respective candidate image and the target image, and wherein the similarity score for each candidate image represents a degree of similarity between the target phenotype and the respective candidate phenotype; determining, by the processor, a threshold similarity score; and determining, by the processor, those candidate images having similarity scores that satisfy the threshold similarity score, wherein the target phenotype is a healthy phenotype, wherein the candidate images having similarity scores that satisfy the threshold similarity score have respective candidate phenotypes corresponding to the healthy phenotype, and wherein those candidate images having similarity scores that do not satisfy the threshold similarity score have respective candidate phenotypes corresponding to an unhealthy phenotype.
-
-
24. A method, comprising:
-
preparing a multi-well sample plate with a target biological cell having a target phenotype and candidate biological cells; applying a variety of candidate treatment regimens to each of the candidate biological cells; recording a target image of the target biological cell; recording candidate images of each of the candidate biological cells, each having a respective candidate phenotype arising in response to the candidate treatment regimen being applied; receiving, by a computing device, the target image and the candidate images; obtaining, by the computing device, a semantic embedding associated with the target image, wherein the semantic embedding associated with the target image is generated using a machine-learned, deep metric network model; obtaining, by the computing device for each candidate image, a semantic embedding associated with the respective candidate image, wherein the semantic embedding associated with the respective candidate image is generated using the machine-learned, deep metric network model; determining, by the computing device, a similarity score for each candidate image, wherein determining the similarity score for a respective candidate image comprises computing, by the computing device, a vector distance in a multi-dimensional space described by the semantic embeddings between the respective candidate image and the target image, and wherein the similarity score for each candidate image represents a degree of similarity between the target phenotype and the respective candidate phenotype; determining, by the computing device, a threshold similarity score; and determining, by the computing device, those candidate images having similarity scores that satisfy the threshold similarity score, wherein the target phenotype is a healthy phenotype, wherein the candidate images having similarity scores that satisfy the threshold similarity score have respective candidate phenotypes corresponding to the healthy phenotype, and wherein those candidate images having similarity scores that do not satisfy the threshold similarity score have respective candidate phenotypes corresponding to an unhealthy phenotype; and selecting, by the computing device, a preferred treatment regimen among the variety of candidate treatment regimens based on the similarity scores.
-
Specification