Simultaneous Mesh Segmentation with Connection Energy Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model-based segmentation methods for multiple objects in image data often result in incorrect spatial relationships and mesh intersections, especially when objects such as internal organs slide relative to each other, as they do not effectively account for spatial interactions between multiple objects.
Innovation Solution
Incorporating a connection energy term that adjusts the spatial relationship between meshes by using a connection transformation, which can include affine transformations and vector differences, to maintain typical distances and prevent mesh intersections, while allowing for sliding displacements and preserving spatial relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If separate adaptation of multiple meshes is performed, then each mesh can be optimized independently, but spatial relationships between objects are not preserved and mesh intersections occur
Solution Approach 1:
The patent merges multiple separate mesh adaptations into a single simultaneous optimization process. The energy function combines individual mesh energies with a connection energy term that enforces spatial relationships: E_total = ΣE_internal(i) + ΣE_external(i) + E_connection. This unified approach ensures that while each mesh adapts to its object, the spatial relationships between meshes are preserved through the connection energy constraint.
2Reliability
If additional edges connecting two meshes are introduced to preserve spatial relationships, then mesh intersection is avoided for pre-positioned objects, but the method fails when objects slide relative to each other
Solution Approach 1:
The patent introduces dynamic connection edges that can change length and position during optimization, unlike the static edges in prior art. The connection energy is formulated as E_connection = Σk_ij * (|e_ij| - |e_ij^ref|)^2, where |e_ij| is the current edge length and |e_ij^ref| is the reference length. This allows the connection constraints to adapt as objects slide, maintaining spatial relationships without preventing natural motion.
Solution Approach 2:
The patent changes the parameters of connecting edges from fixed to variable during optimization. By using a reference configuration |e_ij^ref| established before optimization and comparing it with current configuration |e_ij|, the system allows parameters to change dynamically while maintaining the spatial relationship constraint through the energy minimization process.
3Reliability
If connection energy is added to control spatial relationships, then spatial relationships are maintained, but the complexity of the energy optimization increases
Solution Approach 1:
The patent segments the total energy function into distinct components: E_total = ΣE_internal(i) + ΣE_external(i) + E_connection. Each component has a clear physical meaning and can be computed independently. The connection energy is further segmented into contributions from individual connecting edges, making the overall complex function manageable through modular computation and optimization.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a system (100) for segmenting a plurality of objects in image data using simultaneous model-based image segmentation. A surface mesh is adapted to each object to be segmented. To avoid or reduce the possibility of mesh collision,a plurality of connecting edges for connecting two proximal meshes are used. A connection energy defined for the plurality of connecting edges allows controlling the spatial relationship between the first and second mesh. This is achieved by including in the connection energy expression terms that will increase the connection energy when the lengths of edges of the plurality of connecting edges connecting the first and second mesh decrease. Using the reference configuration of the plurality of connecting edges defined based on the pre- positioned first and second mesh allows taking into account prior knowledge about a typical spatial relationship between the first and second object of the plurality of objects.