Contour Propagation Across Viewframes Using Optical Flow Validation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing AI-based global algorithms for analyzing myocardium mechanics in 2D B-Mode echocardiography images fail to provide correspondence between features and detected boundaries, while local algorithms struggle with tracking speckles when they are not present, leading to inaccurate and noisy tracking across viewframes.

Innovation Solution

Implement an optical flow analysis validated against motion priors to propagate contours across viewframes, using resolved features and motion priors to ensure continuity and accuracy of tracking points, even when speckle patterns are absent.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If AI-based global algorithms are used to detect cardiac boundaries, then boundary detection accuracy is improved, but correspondence between features and detected boundaries is lost

Engineering Contradiction:

Improveboundary detection accuracyVSAvoidfeature correspondence information

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The patent segments the myocardium boundary detection into two independent parts: (1) AI-based global algorithms detect boundaries at key viewframes (end diastolic and end systolic), and (2) local algorithms track speckle patterns across all viewframes. This segmentation allows each method to excel at its strength while the results are combined to provide both accurate boundaries and feature correspondences.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent merges the outputs of AI-based boundary detection and local speckle tracking algorithms into a unified contour tracking system. The AI-detected boundaries serve as constraints and guidance for the local tracking algorithm, combining the strengths of both approaches to achieve both accurate boundary detection and maintained feature correspondence.

Inventive Principle:

Principle #5Merging (Combining)

2Loss of information

If local feature tracking algorithms are used to track speckles, then correspondence between boundary features is improved, but tracking accuracy deteriorates when speckles are absent

Engineering Contradiction:

Improvefeature correspondence informationVSAvoidtracking accuracy

Core Design Contradiction:

Loss of informationVSMeasurement precision

Solution Approach 1:

The patent performs preliminary action by using AI-based algorithms to detect and establish accurate boundary contours at key viewframes (end diastolic and end systolic) before the local tracking algorithm operates. These pre-detected boundaries serve as reliable starting points and constraints that guide the local speckle tracking, ensuring accuracy even when speckles are absent or degraded.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces AI-based boundary detection as an intermediary that bridges the gap between global viewframe analysis and local speckle tracking. This intermediary provides reliable boundary information at key moments, allowing the local tracking algorithm to maintain accuracy without relying solely on potentially absent or degraded speckle patterns.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Stability of the object's composition

If contour tracking is performed across all viewframes, then measurement continuity is improved, but tracking noise increases when speckles are missing

Engineering Contradiction:

Improvemeasurement continuityVSAvoidtracking reliability

Core Design Contradiction:

Stability of the object's compositionVSReliability

Solution Approach 1:

The patent applies local quality by allowing different tracking methods to operate in different regions and time points: AI-based boundary detection is used at key viewframes where global context is most important, while local speckle tracking is used at intermediate viewframes where motion continuity is most important. This differentiated approach optimizes both continuity and reliability across the entire viewing sequence.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS12475575B2Contour motion propagation across viewframe data

Publication Date: 2025.11.18 SIEMENS HEALTHINEERS AG

AI summary

A system includes propagation logic configured to obtain one or more contours for one or more directed viewframes within viewframe data. The one or more contours each having a set of tracking points. The viewframe data further includes intermediate viewframes among the one or more directed viewframes. The propagation logic is configured propagate the one or more contours across the intermediate viewframes via iterative viewframe-to-viewframe propagation. The iterative viewframe-to-viewframe propagation include optical flow analysis to determine candidate locations for tracking points followed by one or more validations using motion priors and/or resolved feature tracking.