Vessel Image Cell Detection for Reliable Monoclonal Seeding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for ensuring cell unity during seeding in well plates, such as those used in antibody drug production, face challenges in accurately distinguishing between true cells and non-cells, leading to potential seeding errors and reduced monoclonality.
Innovation Solution
An information processing device and method that utilizes a trained model to detect cell candidate regions in vessel images, combining deep learning and image processing to enhance sensitivity and accuracy in identifying both cells and cell-like structures, followed by user verification and aggregation of results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detection accuracy is increased to distinguish true cells from non-cells, then measurement precision is improved, but sensitivity is reduced causing true cells to be misclassified as non-cells
Solution Approach 1:
The patent segments the detection task into multiple stages: first detecting all cell-like objects with high sensitivity, then classifying them as true cells or non-cells based on shape characteristics. This segmentation allows the system to maintain high sensitivity in the detection phase while achieving high precision in the classification phase, resolving the contradiction between sensitivity and precision.
Solution Approach 2:
The patent applies different evaluation criteria to different regions of detected objects. Specifically, it evaluates the shape characteristics (circularity, aspect ratio) of detected regions to determine whether they represent true cells or non-cells. This local quality assessment allows the system to maintain high sensitivity while reducing false positives, as different regions are evaluated with appropriate criteria.
2Productivity
If image processing is used to detect cells, then productivity is improved, but measurement precision deteriorates due to difficulty in distinguishing cells from non-cells
Solution Approach 1:
The patent changes the parameters used for cell identification from simple presence/absence detection to shape-based classification. By calculating shape parameters such as circularity and aspect ratio, the system can automatically distinguish true cells from non-cells with high accuracy. This parameter transformation enables both high productivity through automated processing and high measurement precision through shape-based discrimination.
Solution Approach 2:
The patent replaces manual visual inspection with automated image processing that uses shape analysis. Instead of relying on operators to visually distinguish cells from non-cells, the system uses computational algorithms to evaluate shape parameters and automatically classify objects. This substitution maintains high productivity while improving measurement precision through consistent, objective criteria.
3Reliability
If cell detection sensitivity is increased to ensure all cells are detected, then reliability is improved, but measurement precision deteriorates due to increased false positives
Solution Approach 1:
The patent segments the detection and classification processes into distinct stages. The first stage detects all potential cell-like objects with high sensitivity without concern for false positives. The second stage then classifies these detected objects using shape analysis to eliminate false positives. This segmentation allows the system to achieve high sensitivity in detection while maintaining high precision in final identification.
Solution Approach 2:
The patent introduces shape analysis as an intermediary step between detection and final classification. Rather than directly classifying detected objects as cells or non-cells, the system first evaluates their shape characteristics (circularity, aspect ratio) as intermediate features. This intermediary assessment acts as a filter that maintains high sensitivity while reducing false positives through objective shape-based criteria.
Data Source
AI summary
An information processing device detects a cell candidate region for determining a unity of a cell from a vessel image obtained by imaging a vessel in which the cell is seeded and includes at least one processor. The processor performs an acquisition process of acquiring the vessel image, performs a detection process of detecting a cell region including the cell and a cell-like region including an object similar to the cell as the cell candidate regions from the acquired vessel image, and an output process of outputting information indicating the detected cell candidate regions.