3D Reconstruction from Projective Views Using Multi-Object Statistical Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for three-dimensional reconstruction from projective views, such as radiographic images, either provide low-dose two-dimensional information or high-dose three-dimensional information, and existing models are limited to single-object reconstructions, failing to accurately represent overlapping anatomical structures and internal appearances.

Innovation Solution

A method for obtaining a three-dimensional reconstruction from projective views using a multi-object statistical model that incorporates statistical information of shape and appearance, along with relative spatial relationships between objects, allowing for the adjustment of simulated projections to maximize similarity with true projective views, enabling realistic reconstructions of overlapping anatomical structures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If volumetric imaging devices (CT scan, MRI) are used to obtain three-dimensional information, then measurement precision and information completeness are improved, but radiation dose and financial cost increase significantly

Engineering Contradiction:
Improvethree-dimensional information accuracyVSAvoidradiation dose
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent creates a three-dimensional statistical model that copies the essential anatomical structures and their spatial relationships from training data. This virtual model can be adjusted and manipulated without exposing patients to additional radiation, as it is a computational representation rather than a physical imaging process.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent changes the approach from direct physical imaging to statistical parameter modeling. By using statistical models that capture variations in anatomical parameters, the system can generate three-dimensional information through computational adjustment of model parameters rather than through additional physical imaging with radiation.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If subject-specific three-dimensional modeling is performed, then diagnostic precision is improved, but device complexity and processing requirements increase

Engineering Contradiction:
Improvediagnostic precisionVSAvoidmodeling system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary actions by creating a comprehensive three-dimensional statistical model during an offline training phase using training data from multiple subjects. This pre-computed model incorporates statistical variations of anatomical structures, so that during actual diagnosis, only simple parameter adjustments are needed rather than complex real-time modeling.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent inverts the traditional approach by first building a general statistical model from multiple subjects, then adapting it to individual patients through parameter adjustment. Instead of building a custom model for each patient (which would be complex), the system starts with a pre-built statistical model and simplifies the individualization process.

Inventive Principle:
Principle #13The other way round (Inversion)

3Loss of information

If multiple overlapping objects are reconstructed simultaneously, then information completeness is improved, but manufacturing precision and reconstruction accuracy become more difficult to maintain

Engineering Contradiction:
Improveanatomical information completenessVSAvoidreconstruction accuracy
Core Design Contradiction:
Loss of informationVSManufacturing precision

Solution Approach 1:

The patent segments the reconstruction problem by creating separate statistical models for different anatomical structures (e.g., vertebrae, ribs, pelvis) while maintaining their spatial relationships. Each structure can be modeled and adjusted independently within the unified three-dimensional framework, making the complex multi-object reconstruction more manageable and accurate.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent merges multiple individual object models into a unified three-dimensional statistical model that preserves the spatial relationships between overlapping structures. By combining the models within a coordinated framework that enforces anatomical constraints, the system maintains reconstruction accuracy while representing multiple objects simultaneously.

Inventive Principle:
Principle #5Merging (Combining)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This method provides a realistic three-dimensional personalized reconstruction of both shape and appearance for multiple overlapping objects, improving diagnostic accuracy by incorporating detailed anatomical information, including bone density and spatial relationships, which is essential for assessing pathologies and fracture risks.

Implementation Method 1

radiographic projective views obtained by means of X-rays

Methodology Applied
Scientific EffectX-ray: X-Ray

Data Source

PatentEP2534641B1Method for obtaining a three-dimensional reconstruction from one or more projective views and use thereof
Publication Date: 2018.11.28 CETIR CENT MEDIC
  • EP2534641B1 patent drawingFigure 1~2
  • EP2534641B1 patent drawingFigure 3~4

AI summary

Method for obtaining a three-dimensional reconstruction from one or more projective views, and use thereof The method comprises carrying out a 3D/2D registration process for adjusting a three- dimensional multi-object statistical model with statistical information of the shape and appearance of two or more objects and information about relative spatial relationships between different poses of the objects, in order to maximize the similarity between a simulated projection generated from the three-dimensional statistical model, and a true projective view (V). The use consists of using the three-dimensional reconstruction (R) obtained by means of the method for diagnosis.