Efficient training and accuracy improvement of imaging based assay

Patent metadata

Field	Value
Publication number	US-12248533-B2
Application number	US-202318101109-A
Country	US
Kind code	B2
Filing date	Jan 24, 2023
Priority date	Nov 25, 2019
Publication date	Mar 11, 2025
Grant date	Mar 11, 2025

Abstract

The present disclosure relates to devices, apparatus and methods of improving the accuracy of image-based assay, that uses imaging system having uncertainties or deviations (imperfection) compared with an ideal imaging system. One aspect of the present invention is to add the monitoring marks on the sample holder, with at least one of their geometric and/optical properties of the monitoring marks under predetermined and known, and taking images of the sample with the monitoring marks, and train a machine learning model using the images with the monitoring mark.

First claim

What is claimed is:

1. A method of training a machine learning model for an image based assay, wherein a sample in the assay is, during a test, imaged by an imaging system with an imperfection, and wherein the sample contains or is suspected of containing an analyte; comprising:
having the sample forming a thin layer on an imaging area of a sample holder, wherein the sample holder comprises one or more monitoring marks in the imaging area, and wherein at least one of the geometric or optical properties of the one or more monitoring marks is predetermined and known;
imaging, using the imaging system, an original image of the sample on the imaging area of the sample holder;
correcting an imperfection in the original image using the at least one of the geometric or optical properties of the one or more monitoring marks, to generate a corrected image; and
training a machine learning model using the corrected image to generate a trained model for measuring the analyte,
wherein the sample holder comprises:
a first plate and a second plate that are movable to each other in different configurations, including an open configuration and a closed configuration, and the sample is disposed between the first and second plates;
a plurality of spacers on one of the two plates or both, wherein the plurality of spacers are situated between the first and second plates in the closed configuration,
wherein, in the open configuration, the two plates are separated apart, and the spacing between the first and second plates is not regulated by the spacers to facilitate the deposition of the sample on one or both of the first and second plates; and
in the closed configuration that is configured after the sample deposition in the open configuration, at least part of the sample is compressed by the first and second plates into a layer of substantially uniform thickness, and the substantially uniform thickness of the layer is regulated by the two plates and the spacers.

2. The method of claim 1 , wherein correcting the imperfection in the original image to generate the corrected image comprises applying a spatial transform to the original image, wherein the spatial transformation uses a mapping between the positions of the one or more monitoring marks in the original image and the predetermined and known positions of the one or more monitoring marks in the sample holder.

3. A method of an image based assay using a machine learning model, wherein a sample in the assay is, during a test, imaged by an imaging system with an imperfection, and wherein the sample contains or is suspected of containing an analyte, the method comprising:
receiving an original image, imaged by the imaging system, of an imaging area of a sample holder, wherein the imaging area of the sample holder comprises a sample and one or more monitoring marks; wherein at least one of the geometric or optical properties of the one or more monitoring marks is predetermined and known;
correcting an imperfection in the original image using the at least one of the geometric or optical properties of the one or more monitoring marks to generate a corrected image; and
analyzing the analyte using the machine learning model,
wherein the sample holder comprises:
a first plate and a second plate that are movable to each other in different configurations, including an open configuration and a closed configuration, and the sample is disposed between the first and second plates;
a plurality of spacers on one of the two plates or both, wherein the plurality of spacers are situated between the first and second plates in the closed configuration,
wherein, in the open configuration, the two plates are separated apart, and the spacing between the first and second plates is not regulated by the spacers to facilitate the deposition of the sample on one or both of the first and second plates; and
in the closed configuration that is configured after the sample deposition in the open configuration, at least part of the sample is compressed by the first and second plates into a layer of substantially uniform thickness, and the substantially uniform thickness of the layer is regulated by the two plates and the spacers.

4. The method of claim 3 , wherein correcting the imperfection in the original image to generate the corrected image comprises applying a spatial transform to the original image, wherein the spatial transformation uses a mapping between the positions of the one or more monitoring marks in the original image and the predetermined and known positions of the one or more monitoring marks in the sample holder.

5. The method of claim 3 , further comprising:
training a machine learning model using the corrected image to generate a trained model for measuring the analyte.

6. The method of claim 5 , wherein correcting the imperfection in the original image to generate the corrected image comprises applying a spatial transform to the original image, wherein the spatial transformation uses a mapping between the positions of the one or more monitoring marks in the original image and the predetermined and known positions of the one or more monitoring marks in the sample holder.

7. An image-based assay system for analyzing an image of sample, wherein the sample is, during a test, imaged by an imaging system with an imperfection, and wherein the sample contains or is suspected of containing an analyte, the system comprising:
a database system to store images; and
a processing device, communicatively coupled to the database system, to:
receive an original image, captured by the imaging system, of an imaging area of a sample holder, wherein the imaging area of the sample holder comprises a sample and one or more monitoring marks; wherein at least one of the geometric or optical properties of the monitoring marks is predetermined and known;
correct an imperfection in the original image using the at least one of the geometric or optical properties of the one or more monitoring marks, generating a corrected image; and
analyze the corrected image using a machine learning model,
wherein the sample holder comprises:
a first plate and a second plate that are movable to each other in different configurations, including an open configuration and a closed configuration, and the sample is disposed between the first and second plates;
a plurality of spacers on one of the two plates or both, wherein the plurality of spacers are situated between the first and second plates in the closed configuration,
wherein, in the open configuration, the two plates are separated apart, and the spacing between the first and second plates is not regulated by the spacers to facilitate the deposition of the sample on one or both of the first and second plates; and
in the closed configuration that is configured after the sample deposition in the open configuration, at least part of the sample is compressed by the first and second plates into a layer of substantially uniform thickness, and the substantially uniform thickness of the layer is regulated by the two plates and the spacers.

8. The system of claim 7 , wherein a training of the machine learning model uses an image that is corrected using the one or more monitoring marks on a second sample holder which is identical to the sample holder.

9. The system of claim 7 , wherein correcting the imperfection in the original image to generate the corrected image comprises a spatial transform that uses a mapping between the positions of the one or more monitoring marks in the original image and the predetermined and known positions of one or more monitoring marks in the sample holder.

10. The method of claim 2 , wherein the analyte is a cell.

11. The

Assignees

• Essenlix Corp

Inventors

• Li Xing
• Chou Wu
• Chou Stephen Y
• Ding Wei
• Qi Ji

Classifications

• G06N3/094

  Adversarial learning · CPC title

• G06N3/0475

  Generative networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/042

  Knowledge-based neural networks; Logical representations of neural networks · CPC title