Ultrasonic Tumor Segmentation Using Gabor Wavelets and Multi-Scale Level Sets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ultrasonic image segmentation methods struggle with uneven grayscale distribution, high noise, and poor edge continuity, leading to inaccurate tumor edge detection in digestive tract images, particularly due to the limitations of conventional level set algorithms like Mumford-Shah and Chan-Vase models.
Innovation Solution
A Gabor wavelet-fused multi-scale local level set method is employed, utilizing multi-directional Gabor wavelets to enhance images, fuse intermediate results, and construct a level set energy equation with regularization terms to optimize edge detection, iteratively refining the edge position.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If classical segmentation methods such as Canny operator method and thresholding method are used, then the segmentation process is simple, but the edge extraction is inaccurate and incomplete due to uneven grayscale distribution, high noise, and poor edge continuity
Solution Approach 1:
The patent applies multi-scale segmentation by dividing the image processing into different scale levels. The level set method evolves contours at multiple scales to capture edges at various detail levels, allowing accurate edge detection while maintaining computational feasibility through hierarchical processing.
Solution Approach 2:
The patent introduces local region information into the energy functional, allowing the segmentation to adapt to local grayscale variations. By incorporating local intensity statistics and gradient information, the method handles non-uniform grayscale distribution and noise while preserving edge accuracy.
2Productivity
If local area-based level set method with LBF model is used, then reinitialization is not required during iteration, but the algorithm is sensitive to initial contour and has poor robustness to noisy images
Solution Approach 1:
The patent combines multiple energy terms into a composite energy functional that integrates global image information, local region information, and gradient magnitude information. This composite approach creates a more robust segmentation that is less sensitive to noise and initial contour selection while maintaining computational efficiency.
Solution Approach 2:
The patent introduces scale parameters to control the extent of local region analysis. By adjusting the scale parameter, the method adapts to different noise levels and image characteristics, improving robustness without sacrificing computational efficiency.
3Measurement precision
If LRB method is used, then local region information is utilized for segmenting inhomogeneous images, but the computational efficiency is low
Solution Approach 1:
The patent computes local region statistics only within evolving contour regions rather than the entire image. This partial computation approach maintains segmentation accuracy for inhomogeneous images while significantly reducing computational burden compared to full-image processing.
Solution Approach 2:
The patent introduces a scale dimension to the processing, allowing multi-scale analysis that captures both fine and coarse image features. This dimensional extension enables accurate segmentation of inhomogeneous regions while the hierarchical structure improves computational efficiency.
4Measurement precision
If local intensity clustering level set method is used, then segmentation effect is good for images with uniform grayscale variation, but serious over-segmentation occurs for ultrasonic images with weak edges and uneven grayscale variation
Solution Approach 1:
The patent replaces global intensity clustering with local region-based energy terms that adapt to local image characteristics. This allows the method to handle both uniform and non-uniform grayscale variations, and to preserve weak edges in ultrasonic images without over-segmentation.
Solution Approach 2:
The patent makes the energy functional dynamic by incorporating evolving contour information and scale parameters. This dynamic adaptation allows the method to adjust to different image types, maintaining good performance on uniform images while avoiding over-segmentation on ultrasonic images with weak edges.
Data Source
AI summary
Disclosed is a Gabor wavelet-fused multi-scale local level set ultrasonic image segmentation method. In the method, non-uniformity of the grayscale of an ultrasonic image is taken as a texture having cluttered directions, the multi-directional property of Gabor wavelets is used to process the image, and intermediate images in different filtering directions are fused by taking maximum values, so as to obtain an intermediate image having a weakened texture effect and an enhanced difference between a foreground and a background. For the feature of a weak edge of an ultrasonic image, a concept of multi-scale is used to improve the conventional LIC method, Gaussian convolution kernels having different variances are set, and a final edge is obtained by means of average fusion.