Physics-Informed PC-MRI Image Processing for Vessel Morphology
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for analyzing fluid flow in a living body, such as blood flow, suffer from inaccuracies due to insufficient resolution in specifying blood vessel structure and misregistration between PC-MRI and MRA images, leading to inadequate quantification of indicators like wall shear stress (WSS).
Innovation Solution
An image processing device and method using a learning model, specifically physics-informed neural networks (PINNs), trained with a loss function incorporating physical laws, to generate high-resolution flow velocity and morphological images from low-resolution PC-MRI data, enhancing accuracy by aligning with Navier-Stokes equations and boundary conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If MRA image is additionally captured to specify blood vessel structure, then structure specification accuracy is improved, but misregistration between PC-MRI and MRA images occurs leading to reduced analysis accuracy
Solution Approach 1:
The patent extracts the morphological information from the flow velocity image itself rather than relying on separate MRA images. The learning model processes only the PC-MRI flow velocity images to simultaneously obtain both flow velocity data and morphological information, eliminating the need for separate MRA imaging and subsequent registration procedures.
Solution Approach 2:
The patent merges the extraction of flow velocity information and morphological information into a single processing step using one learning model. Instead of separating these into two independent imaging processes (PC-MRI for flow velocity and MRA for morphology), the model integrates both extraction tasks from a single input image, ensuring consistent spatial alignment.
2Measurement precision
If registration between PC-MRI image and MRA image is performed to match blood flow and structure, then spatial matching is improved, but misregistration errors increase reducing quantification accuracy
Solution Approach 1:
The patent extracts morphological information directly from the PC-MRI flow velocity image using the learning model, eliminating the need for separate MRA imaging and subsequent registration. This direct extraction approach removes the source of misregistration errors while maintaining accurate spatial correspondence between flow velocity data and vessel morphology.
Solution Approach 2:
The learning model creates a virtual copy of the morphological information from the flow velocity image data itself. Instead of relying on a separate MRA image that may be misaligned, the model generates a consistent morphological representation copied from the same PC-MRI data, ensuring perfect spatial alignment.
3Measurement precision
If learning model is trained with physics-informed neural networks using loss function with physical laws, then analysis accuracy is improved, but computational complexity increases
Solution Approach 1:
The learning model is pre-trained using physics-informed neural networks with loss functions that incorporate physical laws (Navier-Stokes equations, continuity equations, boundary conditions) before actual analysis. This preliminary training embeds the physical constraints into the model's decision-making process, allowing accurate analysis without requiring complex real-time physical calculations during inference.
Solution Approach 2:
The patent replaces complex real-time physical calculations and iterative solving of fluid dynamics equations with a pre-trained neural network model. Instead of performing computationally intensive CFD simulations during analysis, the model uses learned patterns from training data that incorporates physical laws, significantly reducing computational complexity while maintaining accuracy.
Data Source
AI summary
An image processing device including a processor, wherein the processor is configured to: use a learning model that is a learning model for generating, from a flow velocity image representing a spatial distribution of a flow velocity vector of fluid in a structure of a living body, a flow velocity estimation image having a higher resolution than the flow velocity image and a morphological image representing a morphology of the structure and that is trained based on a loss function using at least information on a physical law of the fluid and the flow velocity image, to generate the flow velocity estimation image and the morphological image from the flow velocity image.