Machine-learnt uncertainty prediction for hemodynamic quantification

Solution Overview

Problem

Current hemodynamic quantification methods for coronary artery disease, such as those using computational fluid dynamics, are computationally demanding and lack real-time capabilities, making them unsuitable for intra-operative guidance, and existing machine learning approaches struggle with uncertainty and sensitivity analysis, which hinders accurate decision-making for physicians.

Innovation Solution

A method and system that utilize machine-learnt classifiers to determine uncertainty, sensitivity, and standard deviation in hemodynamic quantification, allowing for real-time prediction and visualization of these metrics, thereby improving the accuracy and confidence of hemodynamic assessments by focusing on geometric fit and anatomical model precision.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If computational fluid dynamics (CFD) is used for hemodynamic quantification, then prediction accuracy is improved, but computational time and resource requirements increase significantly

Engineering Contradiction:

Improvehemodynamic quantification accuracyVSAvoidcomputational time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent pre-computes sensitivity and uncertainty metrics during the model training phase using CFD simulations on training data. These pre-computed sensitivity patterns are then stored and reused during real-time prediction, eliminating the need for repeated CFD computations during clinical use. This allows accurate hemodynamic quantification with rapid prediction times.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates a simplified machine learning model that copies the essential sensitivity and uncertainty characteristics of the full CFD model. By training the ML model to replicate CFD behavior on training data, it can predict hemodynamic quantities with comparable accuracy but at a fraction of the computational cost during real-time application.

Inventive Principle:

Principle #26Copying

2Productivity

If machine learning approaches are used for hemodynamic quantification, then computational efficiency is improved, but uncertainty and sensitivity analysis capabilities deteriorate

Engineering Contradiction:

Improveprediction speedVSAvoiduncertainty analysis capability

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent pre-computes sensitivity and uncertainty metrics during the model training phase using CFD simulations on training data. These pre-computed sensitivity patterns are then stored and reused during real-time prediction, eliminating the need for repeated CFD computations during clinical use. This allows accurate hemodynamic quantification with rapid prediction times.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent incorporates sensitivity and uncertainty analysis as feedback mechanisms in the ML model architecture. The model not only predicts hemodynamic quantities but also outputs sensitivity scores and uncertainty estimates based on the pre-computed patterns from training data, enabling reliable uncertainty analysis alongside fast predictions.

Inventive Principle:

Principle #23Feedback

3Measurement precision

If detailed anatomical modeling is performed to reduce uncertainty, then measurement precision is improved, but device complexity and processing requirements increase

Engineering Contradiction:

Improvegeometric fit accuracyVSAvoidmodel reconstruction complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies sensitivity analysis to identify specific regions of the coronary anatomy that have the greatest influence on hemodynamic predictions. Instead of uniformly refining the entire anatomical model, the system focuses computational and modeling efforts on high-sensitivity regions, reducing overall model complexity while maintaining prediction accuracy in critical areas.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent pre-computes sensitivity and uncertainty metrics during the model training phase using CFD simulations on training data. These pre-computed sensitivity patterns are then stored and reused during real-time prediction, eliminating the need for repeated CFD computations during clinical use. This allows accurate hemodynamic quantification with rapid prediction times.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS10522253B2Machine-learnt prediction of uncertainty or sensitivity for hemodynamic quantification in medical imaging

Publication Date: 2019.12.31 SIEMENS HEALTHINEERS AG

AI summary

The uncertainty, sensitivity, and/or standard deviation for a patient-specific hemodynamic quantification is determined. The contribution of different information, such as the fit of the geometry at different locations, to the uncertainty or sensitivity is determined. Alternatively or additionally, the amount of contribution of information at one location (e.g., geometric fit at the one location) to uncertainty or sensitivity at other locations is determined. Rather than relying on time consuming statistical analysis for each patient, a machine-learnt classifier is trained to determine the uncertainty, sensitivity, and/or standard deviation for the patient.