Vascular Image Registration Using Intravascular Ground Truth

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Non-invasive vascular imaging, such as CTA, suffers from high inter-user variability in manual delineation of lumen and plaques due to limited resolution and imaging artifacts, hindering effective training and validation of AI/ML models for medical imaging analysis.

Innovation Solution

The registration of intravascular and non-invasive vascular images is achieved by determining a trajectory using a cost function that measures similarity and optimizing rotation angles, allowing for the construction of a 3D mesh and angular registration, utilizing machine learning models to predict unit tangent vectors and simulate the pullback of imaging devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual delineation is used to annotate CTA images, then training data can be obtained, but high inter-user variability occurs due to limited resolution and imaging artifacts

Engineering Contradiction:

Improvedelineation precisionVSAvoidinter-user variability

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent introduces intravascular images as an intermediary reference standard. These images serve as ground truth that mediates between the non-invasive CTA images and the desired precise delineation, allowing automated algorithms to learn from the high-quality intravascular reference data rather than relying on variable manual annotations of lower-quality CTA images

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces manual mechanical annotation processes with automated image registration and AI-based segmentation systems. By using computational methods to register intravascular and non-invasive images, the system eliminates human variability in delineation while maintaining precision through algorithmic consistency

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If intravascular images are used as ground truth, then training accuracy improves, but image registration complexity increases

Engineering Contradiction:

Improvetraining accuracyVSAvoidregistration complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The registration process is segmented into distinct computational steps: extracting fiducial landmarks from both image types, determining correspondence between landmarks, calculating transformation parameters, and applying registration. This segmentation allows each step to be optimized independently and makes the overall complex process more manageable and reproducible

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs preliminary extraction of fiducial landmarks and determination of their correspondence before executing the full registration. This preliminary action simplifies the subsequent registration calculation by providing pre-identified anchor points, reducing the computational complexity of the overall registration task

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentEP4712024A1Registration of intravascular and non-invasive vascular images for training ai models

Publication Date: 2026.03.18 SIEMENS HEALTHINEERS AG

Figure 1
Figure 2
Figure 3

AI summary

Systems and methods for determining a trajectory of one or more intravascular images in one or more non-invasive vascular images are provided. 1) one or more intravascular images of a vessel of a patient and 2) one or more non-invasive vascular images of the vessel of the patient are received. A trajectory of the one or more intravascular images in the one or more non-invasive vascular images is determined. The trajectory is output.