Sequencing Image Compression Using Signal-Noise Pixel Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Gene sequencing images generated by fluorescence or microarray techniques are excessively large and sparse, making them unsuitable for archiving due to the lack of effective compression methods that can maintain clinically relevant information without losing important data.
Innovation Solution
A method that separates pixels into information and noise sets based on peak identification and standard deviation, using a noise threshold parameter to classify pixels, and employs a sparse matrix with a header to store distribution parameters for lossless compression, achieving a compression ratio of 10 to 20 times while preserving essential sequencing data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If common compression techniques are applied to gene sequencing images, then compression ratio is improved, but information loss increases and clinically relevant data is lost
Solution Approach 1:
The patent segments the image data into two distinct sets: a first set of pixels pertaining to information (signal) and a second set of pixels pertaining to noise. This segmentation is achieved by identifying local maxima in the image data and classifying pixels based on their intensity relative to these maxima. By separating signal from noise at the pixel level, the patent enables selective compression that preserves clinically relevant information while reducing overall data size.
Solution Approach 2:
The patent applies different processing quality to different regions of the image. Pixels identified as containing information (local maxima and their neighborhoods) are preserved with high fidelity, while pixels identified as noise are compressed more aggressively. This local quality approach ensures that clinically relevant regions maintain their integrity while non-critical regions contribute less to the total data size.
2Quantity of substance
If direct thresholding is applied to remove noise, then data size is reduced, but spots of lower intensity that might be of clinical relevance are also removed
Solution Approach 1:
The patent changes the parameter used for noise identification from a fixed global threshold to a dynamic local threshold based on the standard deviation of pixel intensities in the neighborhood of each local maximum. This parameter change allows the compression algorithm to adapt to local variations in signal intensity, preserving low-intensity spots that are clinically relevant while still removing true noise. The noise threshold parameter Nth is applied multiplicatively to the local standard deviation, creating a adaptive thresholding mechanism.
3Quantity of substance
If high compression ratios are achieved, then storage space is reduced, but image quality and information integrity deteriorate
Solution Approach 1:
The patent extracts and removes the noise component from the image data before compression. By identifying pixels that belong to the noise set (those within Nth standard deviations of the minimum pixel value) and separating them from the information-containing pixels, the patent achieves high compression ratios by only storing essential information. The noise pixels are either discarded or reconstructed with lower precision, while information pixels are preserved with high fidelity.
Data Source
AI summary
A method for compression and/or decompression of bio-medical images such as DNA sequencing images is provided. The method comprises regeneration of a complete image using multiple distribution functions generated on the fly. The compression ratio achieved is in the order of 10 to 20 times, and is near lossless with respect to the clinically relevant information; which makes it well suited for these DNA sequencing images.