3D Image Segmentation via Synthetic 2D Projection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for segmenting 3D image data sets in medical technology are computationally intensive and often lose relationships between layers, leading to inconsistent results, especially when compared to 2D image data set segmentation methods.

Innovation Solution

A method that generates synthetic 2D image data sets from 3D image data sets using projection techniques, allowing for pre-positioning of a generic model, reducing the 3D data set to a region around the model, and adapting the model to these synthetic 2D data sets to create a precise 3D model of the body structure.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If 3D image data sets are segmented by breaking them down into multiple 2D image data sets, then segmentation can be performed using existing 2D methods, but the relationships between individual layers are lost and results become inconsistent

Engineering Contradiction:
ImproveSegmentation process feasibilityVSAvoidSegmentation consistency
Core Design Contradiction:
Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent applies dimensionality change by projecting 3D volume data onto 2D planes to generate synthetic projection images. This allows the use of efficient 2D segmentation methods while maintaining 3D structural relationships through the projection geometry, thus resolving the contradiction between ease of manufacture and manufacturing precision.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent introduces synthetic projection images as an intermediary between the original 3D volume data and the segmentation process. These projection images serve as a mediator that preserves 3D structural relationships while enabling the application of 2D segmentation techniques, thereby maintaining both feasibility and consistency.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Manufacturing precision

If atlas-based methods are used to segment 3D image data sets directly, then 3D structural relationships are maintained, but the process becomes computationally time-consuming

Engineering Contradiction:
ImproveSegmentation accuracyVSAvoidSegmentation speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent extracts only the essential 3D structural information by projecting volume data onto 2D planes. This extraction reduces the computational complexity from full 3D processing to 2D processing while retaining the critical geometric relationships needed for accurate segmentation, thus improving productivity without sacrificing precision.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent creates synthetic projection images as simplified copies of the 3D volume data from specific viewing angles. These 2D copies are computationally easier to process while still containing sufficient information for accurate segmentation, thereby resolving the contradiction between accuracy and speed.

Inventive Principle:
Principle #26Copying

3Productivity

If 2D segmentation methods are applied to 3D data by slicing, then computational complexity is reduced, but layer relationships are lost

Engineering Contradiction:
ImproveProcessing efficiencyVSAvoidLayer relationship information
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

Instead of slicing 3D data into multiple 2D layers, the patent projects the 3D data onto 2D planes from specific angles. This dimensionality change approach maintains the global structural relationships between different parts of the volume while still enabling efficient 2D processing, thus preventing information loss about layer relationships.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP2083390B1Method for segmenting a 3D image data set, accompanying computer program product and accompanying system
Publication Date: 2016.06.01 BRAINLAB AG
  • EP2083390B1 patent drawingFigure 1

AI summary

The present invention relates to a method for segmenting a 3D image dataset, comprising the following steps: e) providing a 3D image dataset of a body structure to be segmented and a generic model of the body structure; f) generating synthetic 2D image datasets based on the 3D image dataset of the body structure provided in a); g) pre-positioning the generic model with respect to the 3D image dataset of the body structure provided in a); and e) generating a 3D model of the body structure by adapting the generic model to the synthetic 2D image datasets of the body structure generated in step b).