Image Registration Using Negative Landmarks to Prevent Misalignment

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing medical imaging registration algorithms face challenges in accurately aligning images of organs like the lung, where structures may not be in the same position due to differences in patient setup and respiratory state, leading to difficulties in tumor size assessment and diagnosis.

Innovation Solution

The use of negative landmarks, which are not aligned during registration, to improve the accuracy of image alignment by maintaining a specific distance threshold between landmarks, preventing misalignment of non-corresponding structures and enhancing the registration process, especially in challenging situations like CT and 3D ultrasound.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If conventional registration algorithms are used to align images, then the registration process can be automated, but the accuracy of alignment deteriorates when organs are in different positions due to patient setup or respiratory state differences

Engineering Contradiction:

Improveautomation of registration processVSAvoidalignment accuracy

Core Design Contradiction:

Extent of automationVSMeasurement precision

Solution Approach 1:

The patent inverts the conventional landmark approach by using negative landmarks (structures that should NOT be aligned) instead of positive landmarks. The registration algorithm minimizes the distance between negative landmarks in transformed coordinates, thereby pushing non-corresponding structures apart and improving alignment accuracy of corresponding structures automatically

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The patent implements feedback by using the distance measurements between negative landmarks to guide the optimization process. The cost function incorporates these distance measurements, and the registration transformation is iteratively adjusted based on feedback from how well the negative landmark constraints are satisfied

Inventive Principle:

Principle #23Feedback

2Measurement precision

If manual landmark-based registration is used to improve alignment accuracy, then the precision of structure alignment improves, but the complexity of operation increases due to the need to manually review numerous images

Engineering Contradiction:

Improvealignment accuracyVSAvoidoperational simplicity

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The patent enables self-service by allowing the system to automatically identify and process negative landmarks without requiring manual user input. The algorithm autonomously selects anatomical structures that should not be aligned and uses them to guide the registration process, eliminating the need for manual landmark placement while maintaining high accuracy

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

By inverting the landmark concept to use negative landmarks, the system automatically identifies structures that should remain apart, eliminating the need for manual selection of positive landmarks while improving accuracy in challenging registration scenarios

Inventive Principle:

Principle #13The other way round (Inversion)

3Measurement precision

If the registration algorithm aligns all landmarks to minimize similarity measures, then the overall image alignment improves, but non-corresponding structures may be incorrectly aligned due to similar appearance

Engineering Contradiction:

Improveoverall alignment precisionVSAvoidcorrectness of structure matching

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent inverts the conventional approach by using negative landmarks - structures that appear similar but should NOT be aligned. The optimization minimizes distances between negative landmarks in transformed space, which indirectly pushes corresponding structures into correct alignment while preventing misalignment of non-corresponding structures

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The patent converts the harmful effect of similar-looking non-corresponding structures (which cause misalignment) into a beneficial constraint. By identifying these structures as negative landmarks and using their distance relationships as constraints, the system prevents their incorrect alignment while improving overall registration reliability

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Data Source

PatentUS10402991B2Device and method for registration of two images

Publication Date: 2019.09.03 KONINKLIJKE PHILIPS NV

AI summary

The present invention relates to a device and method for registration of two images. To increase the accuracy of the registration the device comprises an input unit (126) for receiving one or more pairs of negative landmarks, which are indicated in the two images, a pair comprising two negative landmarks, one in each image, which structures are not identical and should not be aligned by the registration. Further, a registration unit (128) is provided for performing a registration of the two images by determining a transformation for transforming at least one of the images for registering it with the other image and determining a spacing measure of the distance between a first landmark in the transformed image and a second landmark in the other image, wherein the transformation is determined such that the spacing measure for a first landmark and a second landmark of at least one pair of negative landmarks is above a first distance threshold, whereby the negative landmarks are taken into account in the registration but are not aligned by the registration.