Synthetic Data Classifier for Blood Flow Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating blood flow in patients rely heavily on patient-specific data, which requires a large number of examples for training, is costly, and may not account for outlier conditions due to limited training data availability.
Innovation Solution
A machine-trained classifier is trained using synthetic data, where a computer model is perturbed to generate various examples, and a bench model is altered to measure flow, allowing the classifier to estimate blood flow based on features from medical scan data without relying on patient-specific training data, and accounting for therapeutic alterations and uncertainty.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is used with patient-specific training data, then the classifier can estimate blood flow accurately, but the cost and time to gather sufficient training data increases significantly
Solution Approach 1:
The patent uses synthetic data generated from computer models as copies of real patient data to train the machine learning classifier. This allows the classifier to learn from numerous examples without requiring actual patient data collection, thereby maintaining accuracy while eliminating the time cost of gathering training data.
Solution Approach 2:
The patent transforms the training data source from real patient data to synthetic data by changing the parameter of data origin. This parameter change enables the system to generate unlimited training examples through computational modeling, resolving the contradiction between needing sufficient training examples and the time required to collect them.
2Measurement precision
If machine learning is used with patient-specific training data, then the classifier can estimate blood flow accurately, but the cost to gather sufficient training data increases significantly
Solution Approach 1:
The patent replaces expensive patient-specific training data with computationally generated synthetic data copies. This copying approach maintains the ability to train accurate classifiers while eliminating the substantial costs associated with recruiting, scanning, and processing real patient data.
Solution Approach 2:
The patent uses inexpensive synthetic data generated from computer models as a disposable alternative to expensive patient data. These synthetic examples can be generated unlimited times at low computational cost, replacing the need for costly and limited real patient datasets.
3Productivity
If machine learning is used with limited patient-specific training data, then the classifier can be trained quickly, but outlier conditions are less likely to be accounted for
Solution Approach 1:
The patent creates a dynamic training dataset by generating synthetic data with varied parameters that can adapt to represent different patient conditions including outliers. This dynamic generation allows the classifier to be trained on diverse conditions without being limited by the static constraints of available patient data.
Solution Approach 2:
The patent changes the parameter of data diversity by using synthetic data generation to create examples covering a broad range of conditions including rare outliers. This parameter change enables the classifier to learn from diverse scenarios while maintaining fast training speeds, as synthetic data can be generated rapidly with controlled variability.
Data Source
AI summary
In hemodynamic determination in medical imaging, the classifier is trained from synthetic data rather than relying on training data from other patients. A computer model (in silico) may be perturbed in many different ways to generate many different examples. The flow is calculated for each resulting example. A bench model (in vitro) may similarly be altered in many different ways. The flow is measured for each resulting example. The machine-learnt classifier uses features from medical scan data for a particular patient to estimate the blood flow based on mapping of features to flow learned from the synthetic data. Perturbations or alterations may account for therapy so that the machine-trained classifier may estimate the results of therapeutically altering a patient-specific input feature. Uncertainty may be handled by training the classifier to predict a distribution of possibilities given uncertain input distribution. Combinations of one or more of uncertainty, use of synthetic training data, and therapy prediction may be provided.