Lesion Segmentation Using Directional Statistical Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current imaging systems face challenges in accurately and consistently delineating tumor boundaries, especially for small tumors with blurred boundaries, inhomogeneous lesions, and regions with similar image characteristics, leading to subjective and time-consuming manual processes prone to inter and intra-operator variations.
Innovation Solution
A method and system that utilize a user-selected seed within a structure, derive a directionally dependent statistical model, identify candidate voxels along a radial direction, and segment the structure using these voxels, enabling robust and repeatable lesion boundary delineation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual tumor delineation is performed by experienced physicians, then diagnostic accuracy can be maintained, but the process is time-consuming and subject to inter and intra operator variations
Solution Approach 1:
The system enables automatic tumor segmentation where the algorithm processes the PET image independently without requiring continuous manual intervention. The user only needs to select a seed point inside the tumor, and the system automatically performs the entire segmentation process, eliminating the time-consuming manual delineation while maintaining accuracy through computational methods
Solution Approach 2:
The patent replaces the manual mechanical process of drawing tumor boundaries with an automated computational algorithm. The image processing system uses statistical models and radial direction analysis to automatically identify and segment tumor regions, substituting the physician's manual delineation with an automated digital process that maintains precision without the time cost
2Productivity
If global threshold methods are used for segmentation, then processing speed is improved, but accuracy deteriorates for lesions adjacent to high-uptake regions
Solution Approach 1:
The system transitions from global thresholding to local statistical modeling. Instead of applying a single threshold across the entire image, the algorithm derives directionally dependent statistical models specifically for regions adjacent to the seed point. This local approach allows accurate segmentation of lesions near high-uptake regions by considering local image characteristics rather than global averages
Solution Approach 2:
The patent changes the segmentation parameter from a fixed global threshold to dynamically derived statistical parameters. The system calculates mean and standard deviation values specific to the radial direction and angular position around the seed point, allowing the threshold to adapt to local image conditions and maintain accuracy in challenging regions
3Reliability
If manual segmentation is performed, then operator expertise can be applied, but subjectivity and variability between operators increase
Solution Approach 1:
The system eliminates operator subjectivity by performing segmentation automatically based on mathematical criteria. The algorithm uses statistically derived parameters and radial direction analysis to objectively identify tumor boundaries, removing the variability inherent in manual operator judgment while maintaining reliability through consistent computational application
Solution Approach 2:
The system incorporates feedback through iterative statistical modeling. The algorithm continuously refines the segmentation by analyzing the radial direction and angular position data, adjusting the statistical parameters based on the image characteristics in each region. This feedback mechanism ensures consistent and reliable segmentation results that are not influenced by operator subjectivity
Data Source
AI summary
A method and system for segmenting structures such as lesions in an image is provided. The method comprises selecting one seed inside a lesion in an image either by a user or automatically. The method further includes deriving a directionally statistical model based on a background region or a foreground region of the lesion and determining candidate voxels along a radial direction. The candidate voxels represent the lesion. The method further includes segmenting the lesion using the candidate voxels.