Image Encoding Device for Separating Abnormal and Normal Region Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for medical image analysis struggle to separately treat image features for abnormal regions of interest and normal regions, making it difficult to accurately diagnose and compare medical images.
Innovation Solution
An image encoding device and method that derive and encode both first and second feature amounts to distinguish image features for abnormality and normal regions within a target image, using an encoding learning model to extract and quantify these features, allowing for the reconstruction of image features and label images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If only region of interest detection is performed, then detection speed is improved, but diagnostic accuracy deteriorates due to inability to separately treat abnormal and normal region features
Solution Approach 1:
The patent segments the image feature extraction process into two distinct pathways: one for abnormal regions (region of interest) and one for normal regions. The encoding learning model simultaneously extracts first feature amounts from abnormal regions and second feature amounts from normal regions, allowing separate treatment and analysis of these features without compromising detection speed or diagnostic accuracy.
Solution Approach 2:
The patent introduces an additional dimensional aspect to the feature extraction by deriving not only the first feature amount indicating abnormality but also the second feature amount indicating normal region characteristics. This dual-dimensional approach enables comprehensive diagnostic analysis while maintaining efficient processing through the encoding learning model architecture.
2Productivity
If only similar region search is performed, then search efficiency is improved, but diagnostic completeness deteriorates due to inability to compare both abnormal and normal regions
Solution Approach 1:
The patent segments the similarity search process into two independent search dimensions: searching for images with similar abnormal region features (using first feature amounts) and searching for images with similar normal region features (using second feature amounts). This segmented approach maintains search efficiency while preventing information loss by preserving both types of diagnostic information.
Solution Approach 2:
The patent adds another dimension to the search space by incorporating both first feature amounts (abnormality) and second feature amounts (normality) into the similarity search. This enables comprehensive comparison that considers both pathological and anatomical features, ensuring diagnostic completeness while maintaining search efficiency through the structured feature organization.
3Measurement precision
If feature extraction focuses only on abnormality, then abnormality detection precision is improved, but normal anatomical feature analysis deteriorates
Solution Approach 1:
The patent segments the feature extraction process into two distinct extraction pathways within the encoding learning model: one pathway extracts first feature amounts specifically from abnormal regions to maintain high abnormality detection precision, while another pathway extracts second feature amounts from normal regions to preserve normal anatomical feature information. Both pathways operate simultaneously without interfering with each other.
Solution Approach 2:
The patent introduces another dimension to the feature representation by deriving both first feature amounts (indicating abnormality) and second feature amounts (indicating normality) from the same input image. This dual-feature approach ensures that neither abnormality detection precision nor normal anatomical feature analysis is compromised, as both types of information are preserved and can be analyzed independently.
Data Source
AI summary
A processor encodes a target image to derive at least one first feature amount indicating an image feature for an abnormality of a region of interest included in the target image. In addition, the processor encodes the target image to derive at least one second feature amount indicating an image feature for an image in a case in which the region of interest included in the target image is a normal region.