OCT Image Quality Improvement via Regional Gamma Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional OCT apparatuses face challenges in maintaining high contrast for both retina and non-retina regions during image processing, leading to suboptimal observation of internal structures due to limitations in segmentation and gradation conversion processing, especially in diseased eyes with irregular retinal shapes.
Innovation Solution
An image processing apparatus and method that utilizes a learned model to generate high-quality medical images by improving image quality for different regions, such as the retina, vitreous body, and choroid, through appropriate gradation conversion and segmentation, even in cases of irregular retinal shapes or lesions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If source data with high dynamic range is converted as is into low dynamic range data for display, then the conversion process is simple, but the contrast of the retina portion significantly decreases
Solution Approach 1:
The patent applies parameter changes by performing gradation conversion with different gamma values for different regions. Specifically, it uses gamma conversion with gamma1 for the retina region and gamma2 for non-retina regions, where gamma1 ≠ gamma2. This regional parameter differentiation allows the retina portion to maintain its contrast characteristics while other regions are optimized for their specific display requirements, resolving the contradiction between simple conversion and contrast preservation.
2Manufacturing precision
If low-intensity-side data is discarded to obtain favorable contrast of retina portion, then the contrast of retina is improved, but the contrast of vitreous body portion or choroid portion decreases making observation difficult
Solution Approach 1:
The patent implements local quality by applying different gradation conversion parameters to different regions of the image. The retina region receives processing with gamma1 to optimize its contrast, while non-retina regions (vitreous body and choroid) receive processing with gamma2 to preserve their low-intensity information. This localized approach ensures that each region's specific contrast requirements are met without compromising other regions, thus resolving the contradiction between improving retina contrast and preserving information in other regions.
3Loss of information
If gradation conversion is performed to ensure contrast of vitreous body portion or choroid portion, then observation of internal structure becomes possible, but the contrast of high-intensity retina portion decreases making observation difficult
Solution Approach 1:
The patent resolves this contradiction by applying local quality through region-specific gradation conversion. The non-retina regions (vitreous body and choroid) are processed with gamma2 to enhance their contrast and make internal structures observable, while the retina region is processed with gamma1 to maintain its high-intensity contrast. This spatially differentiated processing allows both regions to achieve their respective optimal contrast levels simultaneously.
4Extent of automation
If segmentation processing determines boundaries utilizing regularity of retinal shape, then automatic detection is achieved, but erroneous detection occurs in diseased eyes with irregular retinal shapes
Solution Approach 1:
The patent applies dynamics by making the segmentation approach adaptive rather than static. Instead of relying solely on fixed regularity-based algorithms that fail for irregular shapes, the system dynamically adjusts by using learned models trained on diverse retinal images including diseased cases. This allows the automatic detection to adapt to varying retinal shapes and conditions, maintaining both automation and reliability even when retinal irregularities are present.
Data Source
AI summary
An image processing apparatus is provided that includes: an obtaining unit configured to obtain a first medical image of a subject; and an image quality improving unit configured to generate a second medical image with image quality higher than image quality of different regions including a first region and a second region that is different from the first region in the obtained first image, using the obtained first image as input data that is input into a learned model.