Microscope Image Analysis Using Preliminary Focal Plane Scanning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis methods for liquids, such as urine, are inefficient in finding the focal plane providing the best contrast and effectively utilizing this information across multiple spatial regions, leading to lengthy and resource-consuming processes.
Innovation Solution
The method involves finding the focal plane with the best contrast in one spatial region and using this information to reduce the depth range that needs to be scanned in other regions, allowing for the selection of images with the best contrast for further processing. Additionally, images taken at different focal planes are used for both training and analysis of image analysis neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the focal plane providing the best contrast is found by scanning through the entire depth range in every spatial region, then the image analysis precision is improved, but the process becomes lengthy and overly resource-consuming
Solution Approach 1:
The patent applies preliminary action by first scanning a preliminary depth range to identify a preliminary best-focused focal plane position before conducting the main image capture. This preliminary scan provides advance information that guides the subsequent focused scanning in each spatial region, avoiding the need to scan the entire depth range from scratch in every region. The preliminary focal plane position determination is performed once and then utilized to inform the scanning strategy for multiple spatial regions, thereby reducing overall process time while maintaining image analysis precision.
2Reliability
If the entire depth range is scanned in every spatial region to find the best contrast image, then the reliability of image analysis is improved, but the resource consumption increases
Solution Approach 1:
The patent applies local quality by adapting the scanning strategy to local conditions in each spatial region. Instead of uniformly scanning the entire depth range in all regions, the system uses the preliminary best-focused focal plane position to determine a localized scanning range for each region. This localized approach focuses computational and scanning resources on the most relevant depth ranges for each specific spatial region, maintaining reliable image analysis while reducing overall resource consumption. The scanning range is locally optimized based on the preliminary findings rather than applying a blanket approach.
3Reliability
If multiple images are taken at different focal planes for training and analysis, then the robustness of image analysis neural network is improved, but the quantity of images required increases
Solution Approach 1:
The patent applies partial action by selecting a limited number of representative images from the captured depth image sequences for training purposes. Rather than using all captured images, the system strategically selects images that provide sufficient training diversity and robustness. The selection is based on the preliminary best-focused focal plane positions and the local scanning ranges, ensuring that the chosen subset of images adequately represents the variation in the data while minimizing the total quantity of images required for training. This selective approach maintains neural network robustness without the overhead of processing excessive images.
Data Source
Figure 1A~1C
Figure 2~3
Figure 4~5
AI summary
The invention is an image taking method comprising: taking images with a microscope in an analyzing space (40) at focal plane positions (50) shifted with an equal step size; and selecting an image from the taken images for further image processing. The images are taken of a sediment (31) of a liquid filled in the analyzing space (40) located between a transparent upper window portion (11) and a transparent lower window portion (21) of a container (30) adapted for analyzing a liquid, wherein the sediment (31) is centrifuged on an inner flat surface of the lower window portion (21 ), and wherein the method comprises: taking, in a first spatial region (41) of the analyzing space (40), a first depth image sequence, and selecting from the first depth image sequence the image with the best contrast for further image processing, and taking, in a second spatial region (42) of the analyzing space (40), a second depth image sequence by taking into account the previous step, wherein the second depth image sequence has fewer images than the first depth image sequence, and selecting from the second depth image sequence the image with the best contrast for further image processing. The invention is further an image analysis method, a method for training an image analysis neural network, and an image analysis neural network based on the above.