Multipass Corner Finding Algorithm for Microarray Image Distortion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-density DNA microarray images present challenges due to non-uniform hybridization, variable chip placement, image distortion, and keystoning effects, making it difficult for current software to accurately identify probe locations and convert image data into biological activity measures without human intervention.
Innovation Solution
A multipass corner finding algorithm using Radon and Fast Fourier transforms to automatically identify microarray chip corners and probes, minimizing scanning errors and correcting keystoning effects, applicable to high-density microarray images with varying resolutions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If high-density microarray chips are used to increase information density, then the quantity of probes increases, but the difficulty of detecting and measuring probe locations increases due to overlapping spots and distortion
Solution Approach 1:
The image processing is divided into multiple sequential passes: first pass identifies corners using Radon transform and FFT, second pass refines corner positions, third pass identifies probe locations. This segmentation of the detection process enables accurate identification of probes even at high densities where spots may overlap
Solution Approach 2:
The patent replaces manual inspection and simple threshold-based software with advanced image processing algorithms including Radon transform and Fast Fourier Transform. This substitution of mechanical/manual methods with sophisticated computational methods enables automatic accurate detection of probe locations in high-density arrays
2Extent of automation
If automated software is developed to identify probe locations, then the extent of automation increases, but the device complexity increases due to need for advanced image processing algorithms
Solution Approach 1:
The algorithm performs preliminary actions by first identifying corner positions using Radon transform and FFT before proceeding to probe location identification. This preliminary structuring of the image data simplifies the subsequent probe detection process and enables full automation
Solution Approach 2:
The multipass algorithm uses feedback mechanisms where the output of one pass (corner positions) becomes the input for the next pass (probe identification). The algorithm refines corner positions iteratively and uses these refined positions to guide probe location detection, creating a self-correcting automated system
3Ease of operation
If current image processing software is used with high-density chips, then the ease of operation is maintained, but the manufacturing precision decreases due to inability to resolve overlapping probes
Solution Approach 1:
The patent changes the processing parameters by applying Radon transform to convert the image to a different domain where linear features become prominent, then uses FFT to identify the dominant orientation. This parameter transformation enables precise detection of probe locations and correction of distortion effects that would be invisible in the original image domain
4Measurement precision
If scanner resolution is increased to improve measurement precision, then the accuracy of image capture improves, but the loss of time increases due to longer scanning duration
Solution Approach 1:
The algorithm performs preliminary analysis using Radon transform and FFT to quickly identify the overall structure, corner positions, and dominant orientations of the microarray. This preliminary processing at lower computational cost enables subsequent detailed probe identification to be performed more efficiently, reducing total processing time
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method accurately identifies corners and probes in high-density microarray images, reducing manual intervention and enabling high-throughput research by accurately converting image data into biological activity measures, with potential to double information density on microarray chips.
Implementation Method 1
applying a Radon transform to an input microarray image to project the image into an angle and distance space where it is possible to find the orientation of the straight lines
Implementation Method 2
applying a fast Fourier transform to the projected image of (a) in order to find the tilting angle of the image
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A method of automatically identifying the microarray chip corners and probes, even if there are no probes at the corners, in a high density and high resolution microarray 5 scanned image having an image space, wherein the method minimizes the error distortions in the image arising in the scanning process by applying to the image a multipass corner finding algorithm comprising: (a) applying a Radon transform to an input microarray image to project the image into an angle and distance space where it is possible to find the orientation of the straight lines; (b) applying a fast Fourier transform to the projected image 10 of (a) to find the optimal tilting angle of the projected image; (c) determining the optimal first and last local maxima for the optimal tilting angle; (d) back projecting the determined first and last local maxima to the image space to find the first approximation of the first and last column lines of the image; (e) rotating the image and repeating steps (a) through (d) to find the first approximation of the top and bottom row lines of the image; (f) determining 15 the first approximation of the four corners of the image from the intersection of the column and row lines; (g) applying a heuristic for determining if the first approximation of step (f) is sufficient; and (h) optionally trimming the scanned image around the first approximation of the four corners and repeating steps (a) through (f).