Digital Pathology Image Compression Using Virtual Optimal Plane
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital pathology systems face significant challenges in storing and transmitting large multi-plane images due to excessive data size, which increases time and cost, and existing methods do not effectively address the redundancy in these images.
Innovation Solution
An image compression method for digital pathology systems that selects optimal blocks with focal points, forms virtual optimal plane images, generates predictive and differential plane images using block descriptors, and applies a Gaussian blur filter to compress the images, reducing redundancy and data size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-plane images are used to capture samples with non-uniform thickness, then image quality and diagnostic accuracy are improved, but data size increases excessively
Solution Approach 1:
The image processing is segmented into multiple stages: optimal block selection from each plane, virtual optimal plane construction, predictive plane generation, and differential image computation. This segmentation allows processing of large multi-plane images in manageable blocks while maintaining overall image quality
Solution Approach 2:
The patent transforms the problem from storing n original planes to storing 1 virtual optimal plane plus n differential planes. This parameter transformation changes the data representation from direct multi-plane storage to a compressed differential form, significantly reducing storage requirements while preserving image quality
2Quantity of substance
If conventional compression algorithms are applied to multi-plane images, then data size is reduced, but image quality deteriorates due to blocking artifacts
Solution Approach 1:
The patent introduces intermediate structures (virtual optimal plane, predictive planes) that serve as mediators between the original multi-plane images and the compressed representation. These intermediaries enable lossless or near-lossless compression by capturing only the differential information needed to reconstruct the original images
Solution Approach 2:
Instead of directly compressing original planes, the patent creates copies in the form of predictive planes that approximate the original images. The difference between originals and predictive copies contains most of the compressible information, enabling efficient compression without quality loss
3Adaptability or versatility
If multiple focal points are used to scan samples with non-uniform thickness, then complete sample coverage is achieved, but storage time and transmission cost increase
Solution Approach 1:
The patent merges information from multiple focal planes by selecting optimal blocks from each plane and combining them into a single virtual optimal plane. This merging process consolidates the essential diagnostic information from all planes into one representative image, reducing storage and transmission requirements while maintaining complete sample coverage
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 significantly reduces the size of multi-plane images while maintaining image quality, as demonstrated by a 4.1 dB improvement in PSNR and reduced blocking artifacts compared to JPEG compression, with recognizable improvements in image fidelity.
Implementation Method 1
The block descriptors use a Gaussian blur filter to form the first to nth original plane images on the basis of the virtual optimal plane image
Data Source
AI summary
Provided is an image compression method of a digital pathology system. The image compression method is a method of compressing digital slide images having first to nth original plane images (n is a natural number greater than or equal to 2). The image compression method includes selecting a block having an optimal focal point as an optimal block from each set of blocks positioned at identical positions of the first to nth original plane images; forming one plane image as a virtual optimal plane image by combining only the optimal blocks; generating block descriptors for forming the first to nth original plane images based on the virtual optimal plane image; generating first to nth predictive plane images from the virtual optimal plane image such that the first to nth predictive plane images are out of focus and most similar to the first to nth original plane images by using location information for the blocks and the block descriptors; generating first to nth differential plane images, the first differential plane image corresponding to a difference between the first original plane image and the first predictive plane image and the nth differential plane image corresponding to a difference between the nth original plane image and the nth predictive plane image; and compressing the first to nth differential plane images.