Multi-Atlas Likelihood Fusion for Brain Segmentation Accuracy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current automated anatomical labeling methods for brain imaging data face challenges in extending generative random diffeomorphic orbit models from single-atlas to multiple-atlas approaches, where joint measurement of parameters such as disease inference, structure volumes, and dense label field estimation are complex due to unknown diffeomorphic changes and simultaneous acquisition of global shape phenotypes.

Innovation Solution

A computer-implemented method that receives imaging data, provides multiple atlases with corresponding candidate regions, co-registers them, assigns probabilities for labeling parameters, and classifies regions of interest based on these probabilities, incorporating a likelihood-fusion approach to integrate information from multiple atlases and account for anatomical variability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple atlases are used for anatomical labeling, then segmentation accuracy is improved, but computational complexity and processing time increase

Engineering Contradiction:

Improvesegmentation accuracyVSAvoidcomputational complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the computational task by processing each atlas separately through co-registration and probability assignment, then combining results through likelihood fusion. This divides the complex multi-atlas problem into manageable independent steps that can be processed sequentially, reducing overall computational complexity while maintaining segmentation accuracy

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces probability maps as intermediary representations between the multiple atlases and the final segmentation result. Each atlas generates a probability map that serves as an intermediate step, allowing the system to fuse information from multiple sources through likelihood combination before producing the final anatomical labeling, thereby managing computational complexity

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If multiple atlases are co-registered and fused, then robustness to anatomical variability is improved, but processing time increases

Engineering Contradiction:

Improverobustness to anatomical variabilityVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent performs preliminary co-registration of multiple atlases to the target image before the actual segmentation process. By pre-aligning the atlases and computing their transformations in advance, the system reduces the computational burden during the main segmentation phase, thereby reducing processing time while maintaining robustness to anatomical variability

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements an iterative expectation-maximization algorithm that continuously refines the probability assignments and label fusion. This continuous refinement process efficiently utilizes computational resources by repeatedly improving the segmentation result without requiring redundant processing steps, thus reducing overall processing time while enhancing robustness

Inventive Principle:

Principle #20Continuity of useful action

3Manufacturing precision

If diffeomorphic transformations are applied for atlas co-registration, then anatomical accuracy is improved, but computational difficulty increases

Engineering Contradiction:

Improveanatomical accuracyVSAvoidcomputational difficulty

Core Design Contradiction:

Manufacturing precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent employs dynamic programming approaches to solve the diffeomorphic transformation problem. By formulating the co-registration as a dynamic optimization problem that evolves through iterative updates of the transformation fields, the system achieves high anatomical accuracy while managing computational difficulty through efficient numerical solutions

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the parameterization of the diffeomorphic transformations by representing them as compositions of simpler transformations with controlled degrees of freedom. This parameter change allows the system to achieve accurate anatomical alignment while reducing the computational complexity of solving the inverse problem, as the transformed parameters are easier to optimize

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS10535133B2Automated anatomical labeling by multi-contrast diffeomorphic probability fusion

Publication Date: 2020.01.14 JOHNS HOPKINS UNIVERSITY

AI summary

A computer-implemented method, system and non-transitory computer readable storage medium for classifying a region of interest of a subject, including receiving imaging data comprising at least one image element, the imaging data comprising the region of interest of the subject; providing a plurality of atlases, each of the plurality of atlases having a candidate region that corresponds to the region of interest of the imaging data, each of the plurality of atlases having at least one image element with associated location and property information; co-registering the plurality of atlases to the imaging data, using at least one processor; assigning a probability to generate a labeling parameter for the region of interest, the probability being associated with each atlas; and classifying the region of interest of the subject based on the assigning.