Cell Image Analysis Using FCN Segmentation and Rotation Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the state and quality of pluripotent stem cells, such as iPS and ES cells, are inefficient and inaccurate due to reliance on human judgment and require selection of specific feature amounts, which can vary with cell species and culture conditions, leading to poor determination efficiency and risk of missing small defective regions or foreign substances.
Innovation Solution
A cell image analysis method using a fully convolutional neural network (FCN) for segmentation, which generates a machine learning model by expanding training data through image rotation and compensation, allowing for accurate identification of undifferentiated and deviated cells without pre-selecting feature amounts, and can detect foreign substances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human judgment is used to determine cell state based on morphological observation, then the determination can be performed without complex processing, but the accuracy is low and determination efficiency is poor
Solution Approach 1:
The patent replaces the mechanical human visual judgment system with an automated image processing system using deep learning convolutional neural networks. The system automatically extracts features from cell images and determines cell states (undifferentiated, deviated, differentiated) without human intervention, thereby improving both accuracy and efficiency simultaneously.
Solution Approach 2:
The patent introduces an intermediate image processing system that acts as a mediator between the raw cell images and the final determination results. This intermediate system uses deep learning models to process images and provide objective, consistent judgments, eliminating the variability inherent in human judgment while maintaining high throughput.
2Ease of operation
If specific feature amounts are selected for cell evaluation, then the processing can be simplified, but the determination accuracy varies with cell species and culture conditions
Solution Approach 1:
The patent develops a universal deep learning-based image processing system that can evaluate multiple types of cells (iPS cells, ES cells, induced cells) under various culture conditions using a single framework. The system automatically adapts to different cell types and conditions without requiring manual selection or adjustment of specific feature amounts, achieving both simplicity and high accuracy across diverse applications.
Solution Approach 2:
The patent employs parameter changes by using deep learning models that can dynamically adjust evaluation parameters based on the input data characteristics. The system automatically learns optimal feature extraction parameters for each cell type and culture condition, eliminating the need for manual parameter selection while maintaining processing simplicity through automated adaptation.
3Ease of operation
If conventional image processing is used for cell evaluation, then the method is straightforward, but small defective regions or foreign substances may be missed
Solution Approach 1:
The patent applies segmentation by dividing the cell image analysis into multiple specialized processing stages: overall cell state evaluation, local region analysis for defects, and foreign substance detection. The deep learning model processes different regions and features separately, ensuring that small defective regions and foreign substances are not overlooked while maintaining overall system simplicity through automated multi-scale analysis.
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
This approach enables quick and accurate non-invasive determination of cell states, improving productivity in cell culture by enhancing identification accuracy and reducing the need for extensive training data preparation, while effectively detecting small foreign substances.
Implementation Method 1
performing segmentation related to a cell by analyzing an observation image of the cell, the cell image analysis method using machine learning as a technique of an image analysis for the segmentation
Data Source
AI summary
A phase image is formed by calculation from a hologram image of a cell, and segmentation is performed for each pixel for the phase image using a fully convolution neural network to identify an undifferentiated cell region, a deviated cell region, a foreign substance region, and the like. When learning, when a learning image included in a mini-batch is read, the image is randomly inverted vertically or horizontally and then is rotated by a random angle. A part that has been lost within the frame by the pre-rotation image is compensated for by a mirror-image inversion with an edge of a post-rotation image as an axis thereof. Learning of a fully convolution neural network is performed using the generated learning image. The same processing is repeated for all mini-batches, and the learning is repeated by a predetermined number of times while shuffling the training data allocated to the mini-batch. The precision of the learning model is thus improved. In addition, since rotationally invariant characteristics can be learned, it is possible to identify cell colonies of various shapes with good precision.