Event-Based Microscopy for Selective Structure Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional microscopy methods are limited by human bias and labor-intensive data analysis, leading to inefficient and time-consuming microscopical studies, especially in high-throughput 3D or 4D imaging tasks.
Innovation Solution
An event-based imaging method using automated microscopes with machine learning algorithms to detect and classify structures of interest, reducing unnecessary image acquisition and enhancing throughput by focusing on regions of interest.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional microscopy methods are used to analyze large specimen volumes, then comprehensive data coverage is achieved, but human bias and labor-intensive analysis reduce throughput and efficiency
Solution Approach 1:
The system performs self-service by automatically acquiring images, detecting structures of interest, classifying them, and generating reports without human intervention. The automated image capturing device coupled with machine learning algorithms enables the system to independently complete the entire analysis workflow, eliminating human bias and labor-intensive manual microscopy while dramatically increasing throughput
Solution Approach 2:
The patent replaces manual mechanical microscopy operations with automated digital systems. Machine learning algorithms substitute human analysts for structure classification and quantification, while automated image capturing devices replace manual specimen examination. This substitution transforms labor-intensive mechanical processes into automated computational workflows, achieving both high throughput and complete automation
2Loss of information
If all regions of specimens are imaged to ensure comprehensive analysis, then complete data coverage is obtained, but data volume and storage requirements increase significantly
Solution Approach 1:
The system extracts only the essential information needed for analysis by using machine learning algorithms to detect and identify structures of interest. Instead of capturing and storing all image data, the system selectively extracts relevant structures, classifies them, and generates concise reports. This extraction approach maintains complete analytical data coverage while dramatically reducing data volume and storage requirements
Solution Approach 2:
The system performs preliminary action by using machine learning algorithms to pre-screen and identify structures of interest before comprehensive analysis. The automated detection and classification processes filter out irrelevant data in advance, ensuring that only meaningful structures are subjected to detailed examination. This preliminary filtering maintains data completeness while minimizing the volume of data requiring storage and further processing
3Productivity
If automated image capturing is used to increase throughput, then productivity improves, but ensuring unbiased and accurate classification becomes more challenging
Solution Approach 1:
The system implements feedback mechanisms where machine learning algorithms automatically classify detected structures, generate reports, and can iteratively improve classification accuracy. The automated classification process provides consistent, unbiased results that are free from human variability. Feedback loops allow the system to learn from classified data and refine its classification accuracy over time, maintaining both high throughput and reliable, unbiased results
Solution Approach 2:
The patent applies parameter changes by utilizing machine learning algorithms with adjustable classification parameters and thresholds. The system can optimize classification accuracy by tuning algorithm parameters based on the specific specimen type and structures of interest. This approach enables automated high-throughput analysis while maintaining reliable and accurate classification through computational parameter optimization rather than human judgment
Data Source
Figure 1
Figure 2a~2d
Figure 3
AI summary
The present invention refers to a computer-implemented method for event- based imaging at least one specimen to record only structures of interest as events, the method comprising: providing an automated image capturing device coupled with a specimen holder, a controller coupled with the image capturing device, at least one processor in an operative conjunction with the controller, and a computer- readable medium comprising instructions that, when executed by the at least one processor, cause the controller and/or the at least one processor to: a) acquire (102), by the image capturing device, at least one initial image of the at least one specimen (100) carried by the specimen holder, b) search (103) the at least one initial image for structures of interest, using an image processing algorithm, C) upon detection of one or more structures of interest, control an imaging capturing software of the image capturing device to acquire (104) at least one main image of the detected structures of interest, respectively, d) classify (105) the detected structures of interest in the at least one main image, using a classifying algorithm, e) evaluate (106) the classified structures of interest, and f) output (107) a result of the evaluated structures of interest, executing the instructions by the at least one processor, wherein at least steps a) to c) are repeated until a pre-given number k of structures of interest has been detected, with k being an integer.