Deformable Model Segmentation for Complex Anatomical Structures

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current medical image processing techniques face challenges in accurately segmenting complex anatomical structures like the apex of the heart or the tip of the liver due to their difficulty in detection within medical images.

Innovation Solution

The use of a deformable model adjusted by known anatomical features and external information, such as stable anatomical structures and distances, to improve segmentation reliability and accuracy, incorporating spring energy and external energy to adapt the model shape.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If deformable models are used for segmentation, then segmentation efficiency is improved, but segmentation reliability deteriorates for complex anatomical structures

Engineering Contradiction:
Improvesegmentation efficiencyVSAvoidsegmentation reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent combines model-based segmentation with image-based segmentation by integrating a deformable model with image gradient information. The deformable model provides initial segmentation efficiency while image gradients provide additional constraints to improve reliability for complex structures like the apex of the left ventricle.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent changes the parameters of the deformable model by incorporating image gradient information as additional constraints. This modifies the energy function of the deformable model to include both model smoothness terms and image gradient alignment terms, improving segmentation reliability without sacrificing efficiency.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If deformable models are used for segmentation, then segmentation speed is improved, but measurement precision deteriorates for difficult-to-detect features

Engineering Contradiction:
Improvesegmentation speedVSAvoidfeature detection precision
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent merges deformable model-based segmentation with gradient-based segmentation. The deformable model provides fast initial segmentation while gradient information provides precise feature location, achieving both speed and precision for difficult-to-detect features.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent uses gradient information as feedback to adjust the deformable model parameters. The gradient-based force acts as a feedback mechanism that continuously refines the segmentation position, improving measurement precision while maintaining the speed advantage of model-based methods.

Inventive Principle:
Principle #23Feedback

3Reliability

If additional anatomical information is incorporated, then segmentation reliability is improved, but device complexity increases

Engineering Contradiction:
Improvesegmentation reliabilityVSAvoidprocessing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies local quality by using gradient information specifically at the boundaries of anatomical structures where it is most needed. Rather than processing the entire image uniformly, the method focuses computational effort on edge regions, improving reliability without proportionally increasing overall complexity.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent performs preliminary action by using the deformable model to generate an initial segmentation before applying gradient-based refinement. This preliminary segmentation provides a good starting point that reduces the computational burden of subsequent refinement steps, managing complexity while improving reliability.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3120323B1Image processing apparatus and method for segmenting a region of interest
Publication Date: 2020.05.20 KONINKLIJKE PHILIPS NV
  • EP3120323B1 patent drawingFigure 1
  • EP3120323B1 patent drawingFigure 2
  • EP3120323B1 patent drawingFigure 3

AI summary

An image processing apparatus (16) is disclosed for segmenting a region of interest (15) in a multi-dimensional image data of an object (12). The image processing apparatus comprises an interface for receiving an image data of the object including the region of interest to be segmented. A selection unit selects a deformable model 30 of an anatomical structure corresponding to the object in the image data. A processing unit segments the region of interest by adapting the deformable model on the basis of the image data (xt) and additional information of the object.