Anatomical Contour Tracing on Low-Resolution Cross-Sectional Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Low-resolution cross-sectional images, such as ultrasound images, fail to precisely visualize the outline of anatomical elements due to disturbances like parasitic echoes, making it difficult to accurately identify and track anatomical features during therapeutic treatments.
Innovation Solution
A method to trace the probable outline of anatomical elements on cross-sectional images by determining a 'seed' pixel and selecting continuous paths with the best score, which can be smoothed and displayed to clearly follow anatomical features, even during movement or deformation, using techniques like dynamic programming and Fourier filtering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cross-sectional imaging is used to visualize anatomical elements, then anatomical structures become visible for treatment planning, but the resolution is too low to precisely delimit the outline of anatomical elements
Solution Approach 1:
The patent applies preliminary action by pre-processing the cross-sectional image to enhance contrast and identify potential anatomical boundaries before the actual contouring process. This includes applying filters to reduce noise and enhance edges, allowing the system to work with improved input data that compensates for the inherently low resolution of the original imaging modality.
Solution Approach 2:
The patent transitions from analyzing only the 2D cross-sectional image to incorporating 3D spatial information by processing multiple sequential slices. By analyzing anatomical structures across multiple dimensions and depths, the system can infer precise contours even when individual 2D slices show poor resolution, effectively adding a dimensional perspective to overcome the resolution limitation.
2Measurement precision
If manual contour tracing is performed on low-resolution images, then anatomical outlines can be identified, but the process is time-consuming and operator-dependent
Solution Approach 1:
The patent implements self-service by enabling the system to automatically perform contour tracing without requiring manual operator intervention. The computer-implemented algorithm autonomously analyzes the cross-sectional image, identifies anatomical structures based on intensity patterns and spatial relationships, and generates precise contours automatically, eliminating the time-consuming manual tracing process while maintaining or improving accuracy.
Solution Approach 2:
The patent applies parameter changes by dynamically adjusting analysis thresholds, intensity ranges, and spatial criteria based on the specific characteristics of each image and anatomical structure. This adaptive parameter adjustment allows the automatic tracing algorithm to optimize its performance for different tissue types and imaging conditions, achieving high accuracy without requiring manual tuning for each case.
3Reliability
If anatomical elements are used as markers for monitoring target movement, then treatment safety can be improved, but the imprecise contours make reliable tracking difficult
Solution Approach 1:
The patent applies preliminary action by pre-establishing reference contours and tracking baselines before treatment begins. By carefully defining the initial anatomical markers and their precise contours in advance using the enhanced image processing methods, the system creates a reliable reference framework that enables accurate movement tracking throughout the treatment procedure, even when real-time image quality varies.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
Method of tracing the likely contour of an anatomical element (S, C) on an image of a cross section of an anatomical site, the image being composed of pixels (P1) having different intensities, the contour of the anatomical element (S, C) being represented by pixels of substantially identical characteristic intensity and bordered by pixels of substantially different intensities so as to visually identify the anatomical element (S, C) on the cross-section image without being able precisely to delimit the contour thereof, the method comprising the following steps: a) forming columns (K) of pixels (Pi) passing through the anatomical element (S, C), b) allocating to each pixel (P1) a pixel score representative of the probability that the pixel is situated on the contour of the anatomical element (S, C), c) constructing a matrix of pixel scores on the basis of several columns (K) of pixels, and defining numerous continuous paths of pixels each passing through a single pixel of each column, d) allocating to each continuous path a path score which is a function of the scores of the pixels constituting the continuous path, e) selecting, from among a set of continuous paths, the path which has the best path score, this path representing the likely contour of the anatomical element.