Cell Image Search Using Multi-Feature Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for searching through time-series cell images struggle to accurately identify desired images among numerous captured images, particularly in distinguishing between images of cell groups maintaining an undifferentiated state and those that have matured or experienced cell liberation, leading to noise in the search results.
Innovation Solution
A cell image search apparatus and method that stores a set of images at different points in time, extracts image feature amounts, and searches for similar images based on these features, ensuring that only images matching multiple criteria in the set are output, thereby reducing noise and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If single image feature amount is used for search, then search speed is improved, but search accuracy deteriorates due to noise from cell liberation
Solution Approach 1:
The patent combines multiple image feature amounts (white streak amount, cell number, cell density) into a composite search criterion. By merging these different features and requiring matches across multiple features rather than relying on a single feature, the system achieves both efficient searching and high accuracy in distinguishing undifferentiated cell groups from liberated cells.
2Measurement precision
If white streak amount is used as search criterion, then undifferentiated cell groups are identified, but noise from cell liberation is included in results
Solution Approach 1:
The patent applies different evaluation criteria to different spatial and temporal contexts. By analyzing white streaks in conjunction with cell number and density changes over time, the system can locally identify undifferentiated states while filtering out liberation events that occur at different stages of cell group development.
Solution Approach 2:
The system uses feedback from multiple image features to validate search results. By requiring that search results satisfy multiple criteria (white streak amount, cell number, density) simultaneously, the system provides mutual verification that eliminates false positives from cell liberation while maintaining sensitivity to undifferentiated states.
3Quantity of substance
If cell number or density is used for search, then images with high cell count are found, but images from high seeding density or weakened cells are included as noise
Solution Approach 1:
The patent performs preliminary analysis of cell group development trajectories by tracking cell number and density changes over time before final search execution. This preliminary action establishes baseline patterns of normal growth versus high seeding density effects, allowing the search to focus on trajectories that match undifferentiated development rather than simply high cell counts.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
There are provided a cell image search apparatus, method, and program capable of searching for a captured image in which a user is interested, among a plurality of captured images obtained by imaging cells in time series, with high accuracy. The cell image search apparatus includes: an image set storage unit (21) that stores an image set including a plurality of cell images at different points in time in a cell culturing process; a captured image acquisition unit (22) that acquires a plurality of captured images obtained by imaging cells in time series; an image feature amount extraction unit (23) that extracts an image feature amount from each cell image of the image set and each of the plurality of captured images; an image search unit (24) that searches for a captured image similar to each cell image of the image set, among the plurality of captured images, based on image feature amounts of each cell image of the image set and the plurality of captured images; and a search result output unit (25) that outputs a search result in a case where captured images similar to at least two cell images included in the image set arc searched for by the image search unit (24).