Cell Monoclonality Image Training Using Optical Virtual Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for ensuring cellular monoclonality in antibody-producing CHO cells require extensive image collection under varied imaging conditions, leading to inefficiencies due to optical virtual images generated by illumination light, which are difficult to distinguish from actual cells.
Innovation Solution
A learning support device and method that generates artificial images depicting optical virtual images caused by illumination light, allowing for efficient training of a machine learning model to distinguish between actual cells and optical artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If captured images are collected under wide variety of imaging conditions to train the machine learning model, then the model training completeness is improved, but the time and effort required for image collection increases significantly
Solution Approach 1:
The patent generates artificial images that copy and simulate optical virtual images (mirror images and convergent images) caused by illumination light, rather than collecting real captured images under various imaging conditions. This allows comprehensive model training without the time-consuming process of actual image collection
Solution Approach 2:
The patent performs preliminary generation of artificial images containing optical virtual images before actual model training. By pre-generating these challenging cases, the system prepares training data in advance, eliminating the need for time-consuming image collection during the training preparation phase
2Measurement precision
If optical virtual images are extracted from actual captured images, then the authenticity of training data is improved, but the complexity and time required for image processing increases
Solution Approach 1:
Instead of extracting optical virtual images from complex actual captured images, the patent generates artificial images that replicate these optical effects through computational methods. This copying approach maintains the authenticity of optical virtual images while avoiding the complexity of extraction from real images
Solution Approach 2:
The patent replaces the mechanical/image processing approach of extracting optical virtual images from captured images with a computational generation approach. This substitution eliminates the need for complex image processing and extraction algorithms
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables the training of a machine learning model to ensure monoclonality without excessive time and effort, improving the accuracy of cell identification in antibody production processes.
Implementation Method 1
an optical virtual image caused by illumination light onto the container is present as an object that should actually be extracted as one cell but is erroneously determined not to be one cell. Specifically, the optical virtual image includes a mirror image that is generated by overlapping a part of one cell
Implementation Method 2
a convergent image that is generated by the illumination light being condensed by a lens effect of the cell
Data Source
AI summary
A learning support device includes a processor, in which the processor is configured to acquire an original image that is a source of a learning input image for a machine learning model used to ensure monoclonality of a cell seeded in a container, and generate, as the learning input image, an artificial image in which an optical virtual image caused by illumination light onto the container is depicted by performing image processing on the original image.