Hemodynamic Prediction Using Recurrent Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting hemodynamic parameters, such as those used in treating coarctation of the aorta and aortic valve disease, rely on invasive catheterization or computationally expensive Computational Fluid Dynamics (CFD) simulations, which are not suitable for clinical practice due to risks and time constraints.
Innovation Solution
A method using an AI system with a Recurrent Neural Network (RNN) like a Long Short Term Memory (LSTM) network to predict hemodynamic parameters by receiving a vessel shape model and flow profile, facilitating faster and more precise predictions compared to traditional methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If CFD simulations are used to compute hemodynamic parameters, then measurement precision is improved, but productivity deteriorates due to computational expense requiring up to one day per scan
Solution Approach 1:
The patent applies preliminary action by pre-computing hemodynamic parameters for a large set of training vessel shapes using CFD simulations before deployment. These pre-computed results are stored in a database, allowing the system to quickly retrieve or interpolate results for new vessel shapes without performing new CFD simulations, thus resolving the contradiction between accuracy and computation speed
Solution Approach 2:
The patent uses copying by creating a comprehensive database of pre-computed hemodynamic parameters for various vessel shapes. Instead of performing new CFD simulations for each patient, the system copies or interpolates results from the pre-computed database, maintaining measurement precision while dramatically improving productivity by reducing computation time from days to minutes
2Measurement precision
If catheterization technique is used to assess hemodynamic parameters, then measurement precision is improved, but object-affected harmful factors increase due to radiation and procedural risks
Solution Approach 1:
The patent introduces an intermediary approach by using patient-specific 3D vessel shape models as a mediator between the patient's anatomy and hemodynamic parameter assessment. Instead of direct invasive measurement, the system creates a digital model of the patient's vasculature and computes hemodynamic parameters from this model, eliminating the need for invasive catheterization while maintaining accuracy
Solution Approach 2:
The patent replaces the mechanical invasive catheterization system with a computational modeling system. By substituting physical intervention with digital simulation based on patient-specific 3D models, the system eliminates radiation exposure and procedural risks while preserving the ability to accurately assess hemodynamic parameters
3Productivity
If standard machine learning approaches are used for hemodynamic prediction, then productivity is improved, but measurement precision deteriorates due to lack of appropriate algorithms for this specific task
Solution Approach 1:
The patent applies parameter changes by transforming the vessel shape into a standardized parameter representation that captures essential geometric features. This parameterization approach, combined with RNN-based sequence processing, allows the system to efficiently process vessel shapes while maintaining the geometric details necessary for accurate hemodynamic parameter prediction, resolving the contradiction between speed and precision
Data Source
Figure 1~3
Figure 4~5
Figure 6
AI summary
The present invention is related to a method of and an Artificial Intelligence (AI) system for predicting hemodynamic parameters for a target vessel, in particular of an aorta, as well as to a computer-implemented method of training an AI unit comprised by said AI system. A vessel shape model of the target vessel and a corresponding flow profile of the target vessel are received. At least one hemodynamic parameter pk is predicted by the AI unit based on the received vessel shape model and the received flow profile. The AI unit is arranged and configured to predict at least one hemodynamic parameter pk based on a received vessel shape model and a received flow profile of the target vessel (aorta).