Vessel Shape Extraction Using ML Landmarks and Path Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying and extracting vessel shapes from CT imaging data are time-consuming and dependent on user expertise, leading to inefficiencies and potential errors in vessel identification.
Innovation Solution
A machine learning model, such as a Deep Q Network, is trained to automatically identify landmarks and determine vessel paths and shapes using radiographic imaging data, minimizing user input by employing pixel intensity thresholds and gradient conditions to segment vessels like the internal carotid and vertebral arteries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual vessel extraction and naming is performed by a user, then vessel identification accuracy can be achieved, but the process is time-consuming and dependent on user expertise
Solution Approach 1:
The system performs automatic vessel extraction, naming, and segmentation using machine learning models and image processing algorithms, eliminating the need for manual user intervention. The computer automatically identifies vessels, extracts their shapes, and names them based on anatomical landmarks and intensity thresholds, making the system self-sufficient and removing the time burden from users.
Solution Approach 2:
The manual mechanical process of user-based vessel extraction is replaced with an automated computational system using deep learning models, gradient-based segmentation algorithms, and pixel intensity analysis. This substitution of manual operation with automated image processing technology resolves the contradiction by maintaining accuracy through algorithmic precision while eliminating time consumption associated with manual work.
2Reliability
If manual vessel extraction is performed, then vessel shape can be identified, but the extraction accuracy is highly dependent on user experience
Solution Approach 1:
The system uses trained machine learning models that automatically perform vessel extraction without relying on user expertise. The algorithms independently analyze image data, identify anatomical landmarks, segment vessels based on intensity thresholds, and name vessels according to predefined anatomical criteria, making the process self-sufficient and eliminating variability introduced by different user skill levels.
Solution Approach 2:
The system transforms the subjective, experience-dependent manual extraction process into an objective, parameter-driven automated process. By using quantifiable parameters such as pixel intensity thresholds, gradient conditions, and machine learning model predictions, the system ensures consistent and reliable extraction accuracy that does not vary with user experience level.
3Productivity
If automatic vessel segmentation is implemented, then user time is reduced, but complexity of the system increases
Solution Approach 1:
The automatic vessel segmentation system is divided into distinct functional modules: a deep Q-network model for landmark identification, a gradient-based algorithm for path determination, an intensity thresholding component for vessel boundary detection, and a naming system based on anatomical rules. This segmentation of the complex automated system into manageable, specialized components enables high processing speed while organizing the inherent complexity into structured, maintainable modules.
4Measurement precision
If manual vessel naming is performed, then anatomical accuracy can be achieved, but the process requires significant user time especially for long and complex vessels
Solution Approach 1:
The system automatically names vessels by identifying anatomical landmarks using the deep Q-network model and applying predefined naming rules based on landmark positions and vessel paths. The computer independently performs the naming task without user intervention, maintaining anatomical accuracy through rule-based identification while eliminating the time required for manual naming, especially for long and complex vessels.
Solution Approach 2:
The system performs preliminary identification of anatomical landmarks and vessel paths using machine learning models before the naming process. By pre-processing the image data to extract vessel geometry and landmark positions, the system prepares all necessary information in advance, enabling rapid and accurate automatic naming without requiring users to spend time on the naming procedure.
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
In an embodiment, a method (100) is described. The method comprises identifying (102) a set of landmarks of a vessel in a subject's body using a machine learning, ML, model configured to identify adjacent landmarks of the vessel from radiographic imaging data of the subject's body. The method further comprises determining (104) a path comprising identified adjacent landmarks in the set. A shape of the vessel between the identified adjacent landmarks is determined based on the determined path and an imaging condition indicative of a wall of the vessel.