Assessment method for simulation models in microscopy

Patent metadata

Field	Value
Publication number	US-12525041-B2
Application number	US-202218559685-A
Country	US
Kind code	B2
Filing date	May 9, 2022
Priority date	May 11, 2021
Publication date	Jan 13, 2026
Grant date	Jan 13, 2026

Abstract

A number of techniques for assessing simulation models for use in microscopy are provided. In one example technique, a first image (IA) of a sample is recorded with a first image recording type, and storing image values of the first image (IA) are stored. Based on a simulation model (SMA→C) being applied to the first image (IA), a simulated image (IA→C) of a third image recording type of the sample is simulated. A third image (IC) of the sample is recorded with a third image recording type. The third image (IC) is compared with the simulated image I(A→C) of the third image recording type for verification of compliance with previously defined quality criteria, and the simulation model (SMA→C) is classified as permissible when the quality criteria are complied with.

First claim

The invention claimed is:

1 . A method for assessing simulation models used in microscopy, the method comprising:
in a first alternative,
recording a first image (I A ) of a sample with a first image recording type;
storing image values of the first image (I A );
predicting, based on a simulation model (SM A→B ) being applied to the first image (I A ), a simulated image (I A→B ) of a second image recording type of the sample;
predicting, based on a simulation model (SM( A→B ) →C ) being applied to the simulated image (I A→B ) of the second image recording type of the sample, a simulated image (I( A→B ) →C ) of a third image recording type of the sample;
recording a third image (I C ) with the third image recording type;
comparing the third image (I C ) with the simulated image (I( A→B ) →C ) of the third image recording type for verification of compliance with previously defined quality criteria; and
classifying the simulation models (SM A→B ) and (SM( A→B ) →C ) as permissible when the quality criteria are complied with;
or, in a second alternative,
recording a first image (I A ) of a sample with a first image recording type;
storing image values of the first image (I A );
predicting, based on a simulation model (SM A→C ) being applied to the first image (I A ), a simulated image (I A→C ) of a third image recording type of the sample;
recording a third image (I C ) with the third image recording type;
comparing the third image (I C ) with the simulated image I( A→C ) of the third image recording type for verification of compliance with previously defined quality criteria; and
classifying the simulation model (SM A→C ) as permissible when the quality criteria are complied with;
or in a third alternative,
recording a first image (I A ) of a sample with a first image recording type;
storing image values of the first image (I A ); predicting, based on a simulation model (SM A→B ) being applied to the first image (I A ), a simulated image (I A→B ) of a second image recording type of the sample, and predicting, based on a simulation model (SM( A→B ) →C ) being applied to the simulated image (I A→B ) of the second image recording type of the sample, a simulated image (I( A→B ) →C ) of a third image recording type of the sample, or predicting, based on a simulation model (SM A→C ) being applied to the first image (I A ), a simulated image (I A→C ) of a third image recording type of the sample, and predicting, based on a simulation model (SM( A→C ) →A ) being applied to the simulated image (I A→C ) of the third image recording type, a simulated image (I( A→C ) →A ) of the first image recording type; comparing the first image (I A ) with the simulated image (I( A→B ) →A ) or with the simulated image (I( A→C ) →A ) for verification of compliance with previously defined quality criteria; and classifying the simulation model (SM A→B ) as permissible when the quality criteria are complied with.

2 . The method of claim 1 , wherein different contrast methods, different channels of a contrast method and/or different illumination powers of a contrast method are used as image recording types.

3 . The method of claim 1 , wherein the simulated images are predicted by means of a machine learning simulation model.

4 . The method of claim 1 , wherein image data of the images (I A ; I C ) in the first and/or third image recording type and/or image data of simulated images are kept as stored data and on request are retrieved and processed.

5 . The method of claim 1 , wherein the quality criteria include previously defined metrics and/or structure-based comparison values.

6 . The method of claim 1 , wherein the quality criteria are assessed by a trained machine learning simulation model.

7 . The method of claim 1 , wherein items of situation-dependent, sample-dependent and/or user-dependent context information are used as further input variables for the verification of the quality criteria.

8 . The method as of claim 1 , wherein a currently attained training state of a model for predicting the respective simulated image subjected to the verification for compliance with the quality criteria is determined depending on a result of the verification.

9 . The method of claim 1 , wherein the first image (I A ) and the third image (I C ) represent the same region of a sample and the image values are registered with respect to one another.

10 . The method of claim 1 , wherein the first, the second and/or the third alternative are/is implemented on the basis of a first image (I A ) of a first location of the sample, and the method further comprising:
predicting a simulated image (I A→B ) based on a simulation model (SM A→B ) being applied to a second image (I′ A ) of a second location of the sample, the captured regions of the first image (I A ) of the first location and of the second image (I′ A ) of the second location not overlapping.

11 . The method of claim 1 , further comprising providing an image predicted by means of a simulation model (SM A→B ) classified as permissible or providing an image (I A→B ) simulated by means of a simulation model (SM A→B ), the prediction of which was classified as permissible.

12 . The method of claim 11 , further comprising:
recoding a second image (I B ) with the second image recording type; storing the second image (I B ); comparing the image (I A→B ) with the stored second image (I B ); and determining differences between the image (I A→B ) and the stored second image (I B ).

Assignees

• Zeiss Carl Microscopy Gmbh

Inventors

• Amthor Manuel
• Haase Daniel

Classifications

• G06T2207/30168

  Image quality inspection · CPC title

• G06T2207/30024

  Cell structures in vitro; Tissue sections in vitro · CPC title

• G06T2207/10056

  Microscopic image · CPC title

• G06T7/0012

  Biomedical image inspection · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title