Blood Vessel Tree ADNN Training with Geodesic Point Grouping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for training artificial deep neural networks to estimate hemodynamic parameters from blood vessel geometry face challenges due to the difficulty in defining universal feature sets, errors in feature selection, and inefficiencies in computational requirements, particularly when using whole or local contexts.
Innovation Solution
A method utilizing an artificial deep neural network (ADNN) with an architecture adapted for point cloud processing, employing geodesic distance-based point grouping to analyze blood vessel trees, allowing for elastic configuration that adapts to fluid dynamics without predefined feature sets, and incorporating computational fluid dynamics for training data generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If CFD simulations are used for hemodynamic parameter estimation, then accuracy of estimation is improved, but computational time and resource requirements increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-training a deep neural network model using CFD simulation data during the development phase. Once trained, the neural network can rapidly estimate hemodynamic parameters for new patient geometries without requiring time-consuming CFD simulations, thus achieving high accuracy while reducing computational time during actual application.
Solution Approach 2:
The patent uses copying by creating a computational surrogate model (neural network) that replicates the results of CFD simulations. The neural network learns the complex relationships between vessel geometry and hemodynamic parameters from CFD training data, then produces similar accurate estimates much faster, effectively copying the high-accuracy functionality of CFD without its computational burden.
2Ease of manufacture
If predefined feature sets are used for neural network training, then training process is simplified, but measurement precision and adaptability to complex vessel geometries deteriorate
Solution Approach 1:
The patent applies segmentation by dividing the complex vessel tree geometry into manageable components through hierarchical point cloud processing. The system segments the vessel tree into multiple levels of detail, processing local regions and global context separately through different network layers, which simplifies the training process while maintaining high precision for complex geometries.
Solution Approach 2:
The patent uses another dimension by transforming 2D image data into 3D point cloud representations of vessel geometries. This dimensional transformation enables the neural network to process spatial relationships and geometric features more effectively, improving measurement precision while the automated extraction process maintains training simplicity.
3Adaptability or versatility
If whole blood vessel tree geometry is used for analysis, then global context is captured, but computational complexity and processing time increase
Solution Approach 1:
The patent applies segmentation by dividing the analysis into hierarchical levels that process different portions of the vessel tree at different scales. Local regions are processed independently in earlier layers, then aggregated to capture global context in later layers, reducing overall processing complexity while maintaining comprehensive global analysis capability.
Solution Approach 2:
The patent uses partial action by processing only the most relevant portions of the vessel tree geometry at each processing stage. The hierarchical approach allows the system to focus computational resources on critical regions while still capturing global context, reducing processing complexity compared to analyzing the entire vessel tree uniformly.
Data Source
Figure 1~2
Figure 3~6
Figure 7
AI summary
A method of training of an artificial deep neural network (ADNN) for estimation of a hemodynamic parameter from a geometry of a blood vessel tree comprising a step of obtaining of a set of geometries of vessel trees, according to the invention is realized with the ADNN having an architecture adapted for point cloud processing with distance based point grouping. The distance is defined as geodesic distance along the blood vessel tree. A method of estimation of hemodynamic parameters from a geometry of a blood vessel tree using an ADNN, according to the invention, involves using ADNN adapted for point cloud processing with geodesic distance based point grouping. The invention concerns also a computer program products comprising a set of instruction that, when run on a computing system, cause it to realize the methods according to the invention. The invention concerns also a computer system adapted to realize methods according to the invention.