Dermoscopic Image Vessel Extraction via Wavelet Transform
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dermoscopic image analysis for distinguishing between benign and malignant tumors is challenging due to the difficulty in accurately extracting clear shape changes and patterns, often resulting in false patterns and decreased diagnostic accuracy.
Innovation Solution
A diagnosis support apparatus and image processing method that separates images into brightness and color components, performs morphology processing to extract candidate regions and likelihoods, and uses smoothing filters to generate vessel-extracted images, reducing false patterns and improving diagnostic accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If morphology processing is applied to extract vessel patterns from dermoscopic images, then vessel extraction capability is improved, but false patterns such as moire artifacts are generated deteriorating diagnostic accuracy
Solution Approach 1:
The patent segments the dermoscopic image into multiple frequency components using wavelet transform. By decomposing the image into different frequency bands, the processing can selectively enhance vessel-related frequency components while suppressing artifacts, thereby extracting vessel patterns accurately without generating false moire patterns.
Solution Approach 2:
The patent changes the processing parameters by using wavelet transform with adjustable decomposition levels and threshold values. By optimizing these parameters, the system can adaptively enhance vessel visibility while minimizing artifact generation, resolving the contradiction between extraction capability and diagnostic reliability.
2Productivity
If top-hat morphology processing is used to extract linear or punctate vessels, then extraction efficiency is improved, but irregular image gradients produce false patterns reducing accuracy
Solution Approach 1:
The patent replaces traditional morphology processing operations with wavelet transform-based processing. This substitution allows for more sophisticated frequency-domain analysis that can handle irregular image gradients without producing false patterns, while maintaining computational efficiency through the properties of wavelet transforms.
3Loss of information
If dermoscopic inspection is performed to eliminate scattered reflection and visualize pigmentation distribution, then observation quality is improved, but clear shape change and pattern extraction remain difficult
Solution Approach 1:
The patent transitions from spatial domain processing to frequency domain processing using wavelet transform. This dimensional change in the processing space allows for better separation of different structural features, making it possible to extract clear shape changes and patterns from the pigmentation distribution data that was previously difficult to analyze.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
First extracting means 101b-1 of a processing unit 101, based on a brightness component and a color information component of a captured image separated by separating means 101a, extract a candidate region using a first morphology processing based on the brightness component, and second extracting means 101b-2 of the processing unit 101 extract a likelihood of a region from a color space composed of the brightness component and the color information component and perform a second morphology processing to generate a region-extracted image, which is displayed on the display device 120. In this case, the morphology processing including the smoothing filter processing is performed on an extracted candidate region and an extracted likelihood of the region.