Atlas-Based Organ Contouring with Shape Priors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional atlas fusion methods in radiation therapy lack smoothness and realism due to the aggregation of propagated masks without shape priors or models, resulting in non-smooth or unrealistic organ contours.
Innovation Solution
A computer-implemented method for atlas-based contouring that constructs a relevant atlas database, selects optimal atlases, propagates and fuses them, and assesses contour quality using deep learning techniques, including similarity learning, affine or deformable mapping, and probabilistic shape models to improve contour accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional atlas fusion aggregates propagated masks by local information without shape priors or models, then the process is simple and fast, but the final results are non-smooth or unrealistic with isolated contours
Solution Approach 1:
The patent changes the parameters of the fusion process by incorporating shape priors and models into the atlas fusion algorithm. This transforms the traditional local information aggregation into a constrained optimization process that enforces anatomical plausibility, thereby improving contour smoothness and realism while managing complexity through structured mathematical formulations.
Solution Approach 2:
The patent introduces shape priors and anatomical models as intermediary constraints between the propagated masks and the final fused result. These intermediaries guide the fusion process to produce anatomically plausible contours, resolving the contradiction between simplicity and quality by adding a layer of anatomical knowledge without requiring complex manual intervention.
2Manufacturing precision
If deep learning techniques and shape models are incorporated into atlas fusion, then contour quality and realism improve, but computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary actions by pre-computing and storing shape priors and anatomical models in the atlas database before the actual fusion process. This allows the fusion algorithm to leverage pre-prepared knowledge structures, reducing real-time computational burden while maintaining high contour accuracy through the use of these pre-processed anatomical constraints.
Solution Approach 2:
The patent applies local quality by using shape priors that are specific to different anatomical regions and organ types. Instead of applying a single global model, the system uses region-specific anatomical knowledge to guide fusion locally, improving accuracy where needed while avoiding unnecessary computational overhead in regions where simple aggregation suffices.
3Measurement precision
If atlases are selected for each organ to accommodate inter-subject variability, then anatomical accuracy improves, but the atlas selection process becomes more complex
Solution Approach 1:
The patent applies segmentation by dividing the atlas selection process into organ-specific components. Instead of selecting a single atlas for the entire volume, the system selects and fuses atlases separately for each organ or anatomical region, accommodating inter-subject variability at the organ level. This segmentation approach improves anatomical accuracy while managing complexity through modular, independent selection processes for each organ.
Data Source
AI summary
Embodiments can provide a method for atlas-based contouring, comprising constructing a relevant atlas database; selecting one or more optimal atlases from the relevant atlas database; propagating one or more atlases; fusing the one or more atlases; and assessing the quality of one or more propagated contours.