IP Library › Granted Patent US 12,731,661
Granted Patent B2
US 12,731,661 · App. 17/596,290 · Granted Sep 8, 2026

System and method for training machine-learning algorithms for processing biology-related data, a microscope and a trained machine learning algorithm

Inventor: Constantin Kappel (Schriesheim, DE)
Assignee: Leica Microsystems CMS GmbH
G16B40/20G06F18/22G06N3/044G06N3/045G06N3/048G06N3/063G06N3/08G16B30/20G16B40/30G06N5/01G06N20/10G06N20/20
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,731,661
App. No.
17/596,290
Granted
Sep 8, 2026
Kind
B2
Abstract

A system ( 100 ) comprises one or more processors ( 110 ) and one or more storage devices ( 120 ), wherein the system ( 100 ) is configured to generate a first high-dimensional representation of the biology-related language-based input training data ( 102 ) by a language recognition machine-learning algorithm executed by the one or more processors ( 110 ). Further, the system ( 100 ) is configured to generate biology-related language-based output training data based on the first high-dimensional representation by the language recognition machine-learning algorithm and adjust the language recognition machine-learning algorithm based on a comparison of the biology-related language-based input training data ( 102 ) and the biolo-gy-related language-based output training data. Additionally, the system ( 100 ) is configured to generate a second high-dimensional representation of the biology-related image-based input training data ( 104 ) by a visual recognition machine-learning algorithm executed by the one or more processors ( 110 ) and adjust the visual recognition machine-learning algorithm based on a comparison of the first high-dimensional representation and the second high-dimensional representation.

Claims (54)

1 . A microscope comprising a system, the system comprising one or more processors and one or more storage devices, wherein the system is configured to:

receive biology-related language-based input training data, wherein the biology-related language-based input training data is a biological sequence;

generate a first high-dimensional representation of the biology-related language-based input training data by a language recognition machine-learning algorithm executed by the one or more processors, wherein the first high-dimensional representation comprises at least three entries each having a different value;

generate biology-related language-based output training data based on the first high-dimensional representation by the language recognition machine-learning algorithm executed by the one or more processors, wherein the biology-related language-based output training data comprises a prediction on a next element in the biological sequence;

adjust the language recognition machine-learning algorithm based on a comparison of the biology-related language-based input training data and the biology-related language-based output training data;

receive biology-related image-based input training data associated with the biology-related language-based input training data from the microscope and generate a second high-dimensional representation of the biology-related image-based input training data by a visual recognition machine-learning algorithm executed by the one or more processors, wherein the second high-dimensional representation comprises at least three entries each having a different value; and

adjust the visual recognition machine-learning algorithm based on a comparison of the first high-dimensional representation and the second high-dimensional representation.

2 . The system of claim 1 , wherein the biology-related image-based input training data is image training data of an image of at least one of a biological structure comprising a nucleotide or a nucleotide sequence, a biological structure comprising a protein or a protein sequence, a biological molecule, a biological tissue, a biological structure with a specific behavior, or a biological structure with a specific biological function or a specific biological activity.

3 . The system of claim 1 , wherein values of one or more entries of the first high-dimensional representation are proportional to a likelihood of a presence of a specific biological function or a specific biological activity.

4 . The system of claim 1 , wherein values of one or more entries of the second high-dimensional representation are proportional to a likelihood of a presence of a specific biological function or a specific biological activity.

5 . The system of claim 1 , wherein the first high-dimensional representation and the second high-dimensional representation are numerical representations.

6 . The system of claim 1 , wherein the first high-dimensional representation and the second high-dimensional representation each comprise more than 100 dimensions.

7 . The system of claim 1 , wherein the first high-dimensional representation is a first vector and the second high-dimensional representation is a second vector.

8 . The system of claim 1 , wherein more than 50% of values of the entries of the first high-dimensional representation and more than 50% of values of the entries of the second high-dimensional representation are non-zero.

9 . The system of claim 1 , wherein the values of more than 5 entries of the first high-dimensional representation are larger than 10% of a largest absolute value of the entries of the first high-dimensional representation and the values of more than 5 entries of the second high-dimensional representation are larger than 10% of a largest absolute value of the entries of the second high-dimensional representation.

10 . The system of claim 1 , wherein the comparison of the biology-related language-based input training data and the biology-related language-based output training data for the adjustment of the language recognition machine-learning algorithm is based on a cross entropy loss function.

11 . The system of claim 1 , wherein the comparison of the first high-dimensional representation and the second high-dimensional representation for the adjustment of the visual recognition machine-learning algorithm is based on a cosine similarity loss function.

12 . The system of claim 1 , wherein the biology-related language-based input training data comprises a length of more than 20 characters.

13 . The system of claim 1 , wherein the adjustment of the language recognition machine-learning algorithm comprises an adjustment of a plurality of language recognition neural network weights, wherein a final set of language recognition neural network weights is stored by the one or more storage devices.

14 . The system of claim 1 , wherein the adjustment of the visual recognition machine-learning algorithm comprises an adjustment of a plurality of visual recognition neural network weights, wherein a final set of visual neural network weights is stored by the one or more storage devices.

15 . The system of claim 1 , wherein the language recognition machine-learning algorithm comprises a language recognition neural network.

16 . The system of claim 15 , wherein the language recognition neural network comprises more than 30 layers.

17 . The system of claim 15 , wherein the language recognition neural network is a recurrent neural network.

18 . The system of claim 15 , wherein the language recognition neural network is a long short-term memory network.

19 . The system of claim 1 , wherein the visual recognition machine-learning algorithm comprises a visual recognition neural network.

20 . The system of claim 19 , wherein the visual recognition neural network comprises more than 30 layers.

21 . The system of claim 19 , wherein the visual recognition neural network is a convolutional neural network or a capsule network.

22 . The system of claim 19 , wherein the visual recognition neural network comprises a plurality of convolution layers and a plurality of pooling layers.

23 . The system of claim 19 , wherein the visual recognition neural network uses a rectified linear unit activation function.

24 . The system of claim 1 , wherein the system is configured to repeat, for each biology-related language-based input training data of a training group of biology-related language-based input training data sets, the steps of:

generating the first high-dimensional representation,

generating the biology-related language-based output training data, and

adjusting the language recognition machine-learning algorithm.

25 . The system of claim 24 , wherein a length of the first biology-related language-based input training data of the training group of biology-related language-based input training data sets differs from a length of second biology-related language-based input training data of the training group of biology-related language-based input training data sets.

26 . The system of claim 1 , wherein the system is configured to repeat, for each biology-related image-based input training data of a training group of biology-related image-based input training data sets, the steps of:

generating a second high-dimensional representation, and

adjusting the visual recognition machine-learning algorithm.

27 . The system of claim 26 , wherein the training group of biology-related language-based input training data sets comprises more entries than the training group of biology-related image-based input training data sets.

28 . A method for training machine-learning algorithms for processing biology-related data using a microscope, the method comprising:

receiving biology-related language-based input training data, wherein the biology-related language-based input training data is a biological sequence;

generating a first high-dimensional representation of the biology-related language-based input training data by a language recognition machine-learning algorithm, wherein the first high-dimensional representation comprises at least three entries each having a different value;

generating biology-related language-based output training data based on the first high-dimensional representation by the language recognition machine-learning algorithm, wherein the biology-related language-based output training data comprises a prediction on a next element in the biological sequence;

adjusting the language recognition machine-learning algorithm based on a comparison of the biology-related language-based input training data and the biology-related language-based output training data;

receiving biology-related image-based input training data associated with the biology-related language-based input training data from the microscope;

generating a second high-dimensional representation of the biology-related image-based input training data by a visual recognition machine-learning algorithm, wherein the second high-dimensional representation comprises at least three entries each having a different value; and

adjusting the visual recognition machine-learning algorithm based on a comparison of the first high-dimensional representation and the second high-dimensional representation.

29 . A non-transitory, computer-readable medium storing a trained machine learning algorithm, the trained machine learning algorithm trained by:

receiving biology-related language-based input training data, wherein the biology-related language-based input training data is a biological sequence;

generating a first high-dimensional representation of the biology-related language-based input training data by a language recognition machine-learning algorithm, wherein the first high-dimensional representation comprises at least three entries each having a different value;

generating biology-related language-based output training data based on the first high-dimensional representation by the language recognition machine-learning algorithm, wherein the biology-related language-based output training data comprises a prediction on a next element in the biological sequence;

adjusting the language recognition machine-learning algorithm based on a comparison of the biology-related language-based input training data and the biology-related language-based output training data;

receiving biology-related image-based input training data associated with the biology-related language-based input training data from a microscope;

generating a second high-dimensional representation of the biology-related image-based input training data by a visual recognition machine-learning algorithm, wherein the second high-dimensional representation comprises at least three entries each having a different value; and

adjusting the visual recognition machine-learning algorithm based on a comparison of the first high-dimensional representation and the second high-dimensional representation.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Apr 24, 2026
From: KAPPEL, CONSTANTIN, MR.
To: LEICA MICROSYSTEMS CMS GMBH
Reel/Frame 074466/0833 →
Continuity (1)
Related Publication 20220246244A1 · Aug 4, 2022
References Cited (17)
US 10127475B1 · Corrado · 2018 [cited by examiner]
US 10134131B1 · Ando et al. · 2018 [cited by applicant]
US 11960518B2 · Kappel · 2024 [cited by examiner]
US 12026191B2 · Kappel · 2024 [cited by examiner]
US 12272161B2 · Kappel · 2025 [cited by examiner]
US 20170249548A1 · Nelson · 2017 [cited by applicant]
US 20240331417A1 · Kappel · 2024 [cited by examiner]
WO 2018091486A1 · 2018 [cited by applicant]
Wang, Jiabao, Learning deep discriminative features based on cosine loss function, Jul. 6, 2017, Electronics Letters, vol. 53, No. 14, p. 918-920 (Year: 2017). [cited by examiner]
Mendieta et al., A cross-modal transfer approach . . . , 2017. [cited by applicant]
Socher et al., Zero-Shot Learning Through Cross-Modal Transfer, Mar. 20, 2013. [cited by applicant]
Frome et al.: “DeViSE: a Deep Visual-Semantic Embedding Model”. [cited by applicant]
Socher et al.: “Zero-Shot Learning Through Cross-Modal Transfer”. [cited by applicant]
Milton et al.: “A cross-modal transfer approach for histological images . . . ”. [cited by applicant]
Yongqin et al.: “Latent Embeddings for Zero-Shot Classification”. [cited by applicant]
Danny et al.: “Embedding Images and Sentences in a Common Space with a Recurrent Capsule Network”. [cited by applicant]
Dai et al.: “Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context”. [cited by applicant]