3D Shape Data Generation via TPS Warp and Normal Line Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating 3-dimensional shape data of organs, such as those used in medical simulators, face challenges including difficulty in setting boundaries from unclear tomographic images and issues with shape transformation accuracy, particularly when normal lines cross, leading to unnatural shapes.
Innovation Solution
A shape data generation apparatus and method that employs a primary transformation using Thin Plate Spline (TPS) Warp to align a reference shape with a target shape by setting landmarks, followed by a secondary transformation where target landmarks are placed internally to avoid crossing normal lines, ensuring a more accurate and natural shape transformation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a predicted shape model is transformed by adding force in the normal direction from the center of gravity of triangular patches, then the shape can be adjusted toward observation data, but when normal lines cross, unnatural shapes occur
Solution Approach 1:
The patent applies preliminary action by performing a first transformation using TPS warp to roughly align the reference shape with the target shape before performing the second transformation. This preliminary alignment ensures that subsequent normal line searches start from a better initial position, reducing the likelihood of crossing normal lines and unnatural shape formation.
Solution Approach 2:
The patent segments the transformation process into two distinct stages: a first transformation using TPS warp with landmark-based control, and a second transformation using normal line search with force addition. This segmentation allows each stage to address specific aspects of shape matching while avoiding the pitfalls of a single monolithic transformation approach.
2Measurement precision
If a doctor draws boundary lines on hundreds of tomographic images to generate 3D shape data, then accurate organ boundaries can be defined, but the workload becomes large
Solution Approach 1:
The patent uses a reference shape as a template or copy that can be transformed to match individual organ shapes. Instead of manually drawing boundaries on each patient's images, the system transforms a pre-existing reference shape to fit the target organ, significantly reducing manual workload while maintaining accuracy through landmark-based alignment.
Solution Approach 2:
The patent transforms the reference shape by changing its geometric parameters through TPS warp and normal line search transformations. By adjusting shape parameters mathematically rather than manually redrawing boundaries, the system achieves accurate boundary definition efficiently without requiring doctors to process hundreds of images individually.
3Ease of manufacture
If transformation is carried out by correlating points in reference shape with points in target shape, then 3D shape data can be generated, but if points are not set properly, transformation cannot be carried out well
Solution Approach 1:
The patent performs preliminary landmark selection and TPS warp transformation before the normal line search. This preliminary action establishes a good initial correspondence between reference and target shape points, making the subsequent transformation process more feasible and accurate without requiring perfect initial point matching.
Solution Approach 2:
The patent introduces TPS warp as an intermediary transformation method that bridges the reference shape and the final transformed shape. This intermediary step with landmark-based control facilitates the point correspondence process, making it easier to achieve accurate transformation even when initial point matching is not perfect.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A disclosed method is a shape data generation method including: identifying, from among a plural vertices of a first shape to be transformed, one or plural first vertices satisfying a predetermined condition including a condition that a normal line of a vertex to be processed crosses with a second shape that is a shape of a transformation target, which is identified from image data; transforming the first shape so as to move each of the one or plural identified first vertices a predetermined distance toward a corresponding normal direction of the identified first vertex; and storing data concerning the plural vertices of the transformed first shape after the identifying and the transforming are executed the predetermined number of times.