Cell Classification via Confidence-Ordered Class Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Pathologists face challenges in efficiently and accurately analyzing multiple overall images for cell classification and counting due to the overwhelming amount of visual information and the need to simultaneously observe cells of different classes, which can lead to errors and inefficiencies in the pathological examination process.
Innovation Solution
An image processing method that uses a computer-assisted algorithm to detect and classify individual cell images, generate class images with cells arranged by confidence levels, and allow users to modify mappings, enabling efficient checking and correction of cell classifications by displaying only relevant portions of class images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If pathologists observe all cells of different classes simultaneously in overall images, then they can acquire comprehensive visual information, but the complexity of visual information increases and efficiency decreases
Solution Approach 1:
The patent segments the visual information by creating separate class images for different cell classes (e.g., mitosis cells, non-mitosis cells). Each class image contains only cells of that specific class, allowing pathologists to observe one cell class at a time without the complexity of mixed classes. This segmentation reduces visual complexity while preserving comprehensive information across all class images.
Solution Approach 2:
The patent extracts cells of the same class from overall images and consolidates them into separate class images. By taking out cells of interest (e.g., all mitosis cells) from the complex overall image context, the system creates simplified views that maintain the extracted information while eliminating distracting elements from other cell classes.
2Measurement precision
If pathologists use optical magnification to observe individual cells, then they can map cells to cell classes accurately, but the time required for analysis increases
Solution Approach 1:
The patent applies preliminary action by using computer-assisted algorithms to pre-detect and classify cells into different classes before pathologist review. The system pre-organizes cells into class images arranged by confidence levels, so that when pathologists review the images, the most reliable classifications are already highlighted and ready for verification, reducing the time needed for manual analysis while maintaining accuracy.
Solution Approach 2:
The patent implements feedback mechanisms where pathologists can correct misclassifications, and these corrections feed back into the system to improve future automatic classifications. The confidence level ordering provides immediate feedback on classification reliability, allowing pathologists to focus their time on lower-confidence cases while trusting high-confidence automated classifications.
3Reliability
If pathologists review all cell images from multiple overall images, then they can verify classification results thoroughly, but the workload and time required increase significantly
Solution Approach 1:
The patent applies local quality by organizing class images according to confidence levels, with higher-confidence classifications positioned for优先 review. This creates a differentiated viewing experience where pathologists can quickly verify the most reliable classifications while dedicating more time to lower-confidence cases, thereby maintaining thorough verification without uniformly increasing workload across all images.
Solution Approach 2:
The patent enables partial action by allowing pathologists to review only the portions of class images that require verification, rather than forcing review of all cell images. The confidence-level ordering system allows selective review of critical cases, maintaining adequate verification reliability while significantly improving productivity by avoiding unnecessary review of high-confidence classifications.
Data Source
AI summary
An image processing method is provided for displaying cells from a plurality of pathology images or overall images. A respective overall image represents a respective patient tissue sample or a respective patient cell sample. The method includes the steps of: providing the overall images, detecting individual cell images in the overall images by means of a computer-assisted algorithm, determining classification data by means of the computer-assisted algorithm, wherein the classification data indicate a respective unique mapping of a respective detected cell image to one of a plurality of cell classes, and wherein the classification data further have a respective measure of confidence in respect of the respective unique mapping, generating respective class images for the respective cell classes, wherein a class image of a cell class reproduces the cell images mapped to the cell class in a regular arrangement and with a predetermined order, and wherein further the order of the mapped cell images is chosen on the basis of the measures of confidence of the mapped cell images, and further, displaying a portion of at least one class image.