Bright-Field Cell Region Classification with Pseudo-Reversed Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for cell observation in bright field imaging struggle with inconsistent contrast states, leading to reduced accuracy in cell region extraction due to variations in image focus, which complicates the collection of uniform training images for machine learning models.
Innovation Solution
A method to generate pseudo-reversed images based on original images to create training data with symmetrical contrast states, allowing machine learning to construct a classification model capable of accurately extracting cell regions regardless of the original image's contrast state.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If images are collected from past libraries without uniform contrast states, then workload is reduced, but classification model accuracy deteriorates
Solution Approach 1:
The patent generates pseudo-reversed images by inverting the luminance of original images to create synthetic training data. This copying approach allows the system to use existing past images while generating additional contrast variations, thereby maintaining workload reduction while improving model accuracy through diverse training samples
Solution Approach 2:
The patent applies luminance inversion to transform images between first and second contrast states. By changing the luminance parameter of existing images, the system creates training data with uniform contrast distribution without requiring new imaging, thus resolving the contradiction between workload and accuracy
2Ease of operation
If images with mixed contrast states are used for training, then data collection is simplified, but extraction accuracy deteriorates
Solution Approach 1:
The patent systematically transforms images with mixed contrast states into a unified contrast state through luminance inversion. By detecting and correcting contrast state parameters, the system maintains data collection simplicity while ensuring uniform contrast distribution for accurate extraction
Solution Approach 2:
The patent inverts the luminance of images to convert between first and second contrast states. This inversion approach allows flexible transformation of mixed contrast data into uniform training data, preserving ease of operation while improving extraction accuracy
3Measurement precision
If new imaging is performed to collect uniform contrast state images, then model accuracy is improved, but workload increases enormously
Solution Approach 1:
Instead of performing new imaging, the patent generates pseudo-reversed images by inverting luminance of existing images. This copying strategy creates uniform contrast training data without additional imaging workload, thereby improving model accuracy while maintaining productivity
Solution Approach 2:
The patent replaces the mechanical process of new imaging with a computational luminance inversion process. This substitution eliminates the need for physical re-imaging while achieving uniform contrast training data, resolving the contradiction between accuracy and workload
Data Source
Figure 1
Figure 2
Figure 3A~3C
AI summary
An image processing method according to the invention includes obtaining a ground truth image teaching a cell region occupied by a cell in an original image for each of a plurality of the original images obtained by bright-field imaging of the cell, generating a reverse image by reversing luminance of the original image at least for the cell region based on each original image, and constructing a classification model by performing machine learning using a set of the original image and the ground truth image corresponding to the original image and a set of the reverse image and the ground truth image corresponding to the original image as a basis of the reverse image respectively as training data.