Tumor Contour Retrieval Using Reference Segments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for retrieving tumor contours from ultrasonographic images, such as the Snake algorithm, face challenges with indistinct boundaries and require manual delineation, leading to time-consuming and inaccurate results, especially for images with low contrast or unclear boundaries.
Innovation Solution
A method for retrieving tumor contours in an image processing system that involves defining a tumor contour annular region and reference segments, retrieving suggestion points using various methods (moving variance, contrast, distance, gradient EWMA difference, and angle), and linking these points to form the contour, allowing for rapid and accurate identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual delineation is used to determine tumor contour, then the contour can be accurately defined by expert knowledge, but the process is time-consuming and subject to inter-observer variation
Solution Approach 1:
The patent divides the tumor contour determination into multiple reference segments across the annular region, with each segment containing multiple measured points. This segmentation allows automated processing of multiple local regions independently, reducing overall processing time while maintaining accuracy through systematic sampling of the contour boundary.
Solution Approach 2:
The patent replaces manual mechanical delineation by doctors with an automated computer-based system that uses mathematical algorithms. The system automatically processes ultrasonographic images, calculates reference segments, and determines contour points without human intervention, thereby eliminating time loss and inter-observer variation while maintaining measurement precision through computational methods.
2Extent of automation
If Snake algorithm is applied to retrieve tumor contour, then automated contour extraction is achieved, but the algorithm fails when boundaries are indistinct or contrast is low
Solution Approach 1:
The patent applies different processing strategies to different local regions (reference segments) based on their specific characteristics. Each reference segment contains multiple measured points that are evaluated independently, allowing the system to adapt to local variations in image quality and boundary clarity, thereby maintaining reliability across diverse ultrasonographic images.
Solution Approach 2:
The patent performs preliminary actions by pre-defining reference segments and measuring multiple points along the tumor contour before applying the contour retrieval algorithm. This preliminary preparation creates a structured framework that guides the automated extraction process, ensuring reliable results even when boundaries are indistinct by having pre-established reference structures to anchor the contour determination.
3Measurement precision
If multiple reference segments with multiple measured points are used, then contour accuracy is improved, but calculation complexity increases
Solution Approach 1:
The patent segments the tumor contour into multiple reference segments, each containing multiple measured points. This segmentation strategy improves accuracy by systematically sampling the contour boundary at multiple locations, while the modular structure of processing each segment independently helps manage computational complexity through divide-and-conquer approach.
Solution Approach 2:
The patent uses multiple measured points (excessive sampling) along each reference segment to ensure accurate contour representation. By measuring more points than strictly minimum required, the system compensates for potential measurement errors and image quality variations, thereby improving contour similarity to the actual contour through redundant sampling.
Data Source
AI summary
The present invention related to a method for retrieving a tumor contour of an image processing system that includes a memory storing a grayscale image and a processor, comprising: receiving an input tumor contour of the grayscale image; defining a tumor contour annular region and a plurality of reference segments of the grayscale image, wherein the input tumor contour is in the tumor contour annular region, and each of the plurality of reference segments is across the tumor contour annular region and includes a plurality of measured points; retrieving a tumor contour suggestion point on each of the plurality of reference segments; and linking all the tumor contour suggestion points on all of the reference segments, for forming the tumor contour. Accordingly, by applying the method of the present invention, a doctor can rapidly and accurately identify the contour of a tumor in a grayscale image.