# US 12,602,940 B2 · Single cell identification for cell sorting

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,602,940 B2
- Document ID: US12602940B2
- Status: Granted
- Kind: B2
- Application number: 18/071,943
- Filed: 2022-11-30
- Published: 2024-05-30
- Granted: 2026-04-14
- Earliest claimed: 2022-11-30
- Assignees: Sony Group Corporation; Sony Corporation of America
- Applicants: Sony Group Corporation; Sony Corporation of America
- Inventors: Haipeng Tang; Michael Zordan; Ming-Chang Liu
- Web page: https://patentgrep.com/patent/US12602940B2

## CPC

- G06V 20/698: {Matching; Classification}
- G06F 18/2431: Multiple classes
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06V 10/44: Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V 10/764: using classification, e.g. of video objects
- G06V 10/82: using neural networks
- G06V 20/695: {Preprocessing, e.g. image segmentation}
- G06V 20/70: Labelling scene content, e.g. deriving syntactic or semantic representations
- G06V 2201/03: Recognition of patterns in medical or anatomical images
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The single cell identification described herein utilizes cell image information and extracts cell features with a neural network model to subtly distinguish the noise events from single cells, allowing the user to choose which different types of noise events to exclude depending on the requirement of applications. The fast neural network model is able to extract more abundant and specific cell features than handpicked features, which enables the model to be equipped with higher accuracy and higher discriminative capability of distinguishing noise events and identifying the single cells in real-time. Utilization of a neural network model for real-time single cell identification represents a novel technique never applied before. It allows high discriminative capability and high accuracy compared to traditional FACS (Fluorescence-activated Cell Sorting). The usefulness of this technique is to integrate with any brightfield (BF) model and fluorescence (FL) model to identify single cells for different downstream applications.

## Claims

### Claim 1

A method programmed in a non-transitory memory of a device comprising:
receiving input;
processing the input using a neural network to identify a noise event;
processing the input using the neural network to identify a single cell, wherein the neural network comprises two convolutional layers and a fully connected layer, wherein the neural network comprises a plurality of kernels to extract features with different fine-grained levels from the input, wherein the neural network is configured to classify more than 2,000 images per second;
classifying the input as the noise event based on processing the input;
classifying the input as the single cell based on processing the input, wherein classifying the input as the single cell based on processing the input includes bleb events and out-of-focus events when an increased target cell amount is determined; and
implementing a brightfield model or a fluorescence model in parallel with classifying the input.

### Claim 2 (depends on 1)

The method of claim 1 wherein the input comprises a plurality of cell images.

### Claim 3 (depends on 1)

The method of claim 1 wherein the noise event is selected from the group consisting of aggregate, debris, doublet, edge, out-of-focus, detached-bleb, and bleb.

### Claim 4 (depends on 1)

The method of claim 1 wherein classifying the input comprises labeling each image with the noise event or the single cell.

### Claim 5 (depends on 1)

The method of claim 1 wherein when the input is classified as the noise event in a classification of out-of-focus, detached-bleb or bleb, the input is retained for further analysis including cell sorting.

### Claim 6

An apparatus comprising:
a non-transitory memory for storing an application, the application for:
receiving input;
processing the input using a neural network to identify a noise event;
processing the input using the neural network to identify a single cell, wherein the neural network comprises two convolutional layers and a fully connected layer, wherein the neural network comprises a plurality of kernels to extract features with different fine-grained levels from the input, wherein the neural network is configured to classify more than 2,000 images per second; and
classifying the input as the noise event based on processing the input;
classifying the input as the single cell based on processing the input; and
implementing a brightfield model or a fluorescence model in parallel with classifying the input; and
a processor coupled to the memory, the processor configured for processing the application.

### Claim 7 (depends on 6)

The apparatus of claim 6 wherein the input comprises a plurality of cell images.

### Claim 8 (depends on 6)

The apparatus of claim 6 wherein the noise event is selected from the group consisting of aggregate, debris, doublet, edge, out-of-focus, detached-bleb, and bleb.

### Claim 9 (depends on 6)

The apparatus of claim 6 wherein classifying the input comprises labeling each image with the noise event or the single cell.

### Claim 10

A system comprising:
a microscope configured for acquiring a cell image; and
a computer configured for:
receiving the cell image;
processing the cell image using a neural network to identify a noise event;
processing the input using the neural network to identify a single cell, wherein the neural network comprises two convolutional layers and a fully connected layer, wherein the neural network comprises a plurality of kernels to extract features with different fine-grained levels from the input, wherein the neural network is configured to classify more than 2,000 images per second;
classifying the cell image as the noise event based on processing the cell image;
classifying the cell image as the single cell based on processing the cell image; and
implementing a brightfield model or a fluorescence model in parallel with classifying the input.

### Claim 11 (depends on 10)

The system of claim 10 wherein the noise event is selected from the group consisting of aggregate, debris, doublet, edge, out-of-focus, detached-bleb, and bleb.

### Claim 12 (depends on 10)

The system of claim 10 wherein classifying the cell image comprises labeling the cell image with the noise event or the single cell.