Machine-Learning Cell Culture Control for Uniform Fluidic Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cell culture systems using multiple fluidic devices struggle to maintain homogeneous cell states due to variations caused by the disposition position of each device, despite efforts to maintain uniform culture environments.
Innovation Solution
An information processing apparatus and method that utilizes machine learning to generate a learned model based on evaluation values and disposition positions, adjusting culture environments to ensure uniformity across multiple fluidic devices using a deep neural network and backpropagation method.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple fluidic devices are arranged in a closed environment such as an incubator, then cell culture can be performed in parallel, but the states of cells vary depending on the disposition position of each fluidic device
Solution Approach 1:
The system applies local quality by individually adjusting the culture environment parameters (temperature, CO2 concentration, humidity) for each fluidic device based on its specific disposition position. The adjustment unit modifies environment parameters locally for each device to compensate for position-dependent variations, ensuring that cells at different positions achieve uniform states despite environmental gradients within the incubator.
Solution Approach 2:
The invention changes physical parameters of the culture environment (temperature, CO2 concentration, humidity) for each fluidic device based on its disposition position. The adjustment unit varies these parameters locally to compensate for positional effects, transforming the uniform parameter setting into position-specific parameter optimization to achieve homogeneous cell states across all devices.
2Ease of operation
If the culture environment is uniformly set for all fluidic devices, then system operation is simplified, but cell states still vary due to disposition position differences
Solution Approach 1:
The system transitions from a static, uniform culture environment setting to a dynamic, position-adaptive control system. The adjustment unit continuously or periodically modifies culture environment parameters based on the disposition position of each fluidic device, enabling the system to adapt to positional variations while maintaining automated control through the learning model that predicts optimal parameters for each position.
3Manufacturing precision
If manual adjustment of culture environment is performed for each fluidic device, then cell state uniformity can be improved, but system complexity and operation difficulty increase
Solution Approach 1:
The system implements self-service through the learning model that automatically learns the relationship between disposition position and optimal culture environment parameters. Instead of requiring manual expert adjustment, the system autonomously determines the appropriate culture environment settings for each fluidic device based on its position, eliminating the need for complex manual intervention while achieving uniform cell states.
Solution Approach 2:
The invention incorporates feedback mechanisms where the states of cells in each fluidic device are evaluated, and this information is fed back to the adjustment unit. The learning model uses this feedback to refine its predictions of optimal culture environment parameters for different disposition positions, creating a closed-loop control system that automatically maintains cell state uniformity without increasing operational complexity.
Data Source
AI summary
An information processing apparatus includes an acquisition unit that acquires an evaluation value for a state of cells which are cultured in each of a plurality of fluidic devices, culture environment, and information representing a disposition position of the fluidic device; and a generation unit that generates a learned model in which the evaluation value and the information representing the disposition position are received as an input and the culture environment is output, through machine learning using, as training data, the evaluation value before adjustment of the culture environment, the culture environment after the adjustment, and the information representing the disposition position, in a case where, due to the adjustment, the evaluation value becomes a first threshold value or more and an absolute value of a difference in evaluation value between the plurality of fluidic devices becomes a second threshold value or less.