Brain Image Normalization via Landmark Segmentation

Solution Overview

Problem

Existing methods for brain image normalization fail to accurately normalize brain images and anatomical structures in the vicinity of the brain, such as the eyes, diencephalon, fornix, corpus callosum, left hippocampus, and right hippocampus, which are crucial for effective discrimination between normal and dementia-related brain images.

Innovation Solution

A brain image normalization apparatus that detects at least four reference landmarks (left eye, right eye, diencephalon, fornix, corpus callosum, left hippocampus, and right hippocampus) using template matching and registration units, performing similarity or nonlinear transformations to normalize the brain images based on these landmarks, and optionally includes local landmarks for further registration and normalization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Shape

If the entire brain image is registrated with a standard brain image using affine transformation, then the shape of the brain image can be normalized, but the anatomical structures included in the vicinity of the brain cannot be normalized

Engineering Contradiction:

Improvebrain shape normalizationVSAvoidanatomical structure normalization precision

Core Design Contradiction:

ShapeVSManufacturing precision

Solution Approach 1:

The patent segments the normalization process into two distinct stages: first performing affine transformation on the entire brain image to achieve basic shape normalization, then performing non-linear transformation on specific anatomical structures (eyes, diencephalon, fornix, corpus callosum, hippocampi) to achieve precise anatomical normalization. This segmentation allows each transformation method to be optimized for its specific target, resolving the contradiction between brain shape normalization and anatomical structure normalization precision.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies different transformation qualities to different regions: global affine transformation for the overall brain shape and local non-linear transformation for specific anatomical structures. This local quality approach ensures that each region receives the appropriate level and type of normalization, achieving both brain shape normalization and anatomical structure normalization precision simultaneously.

Inventive Principle:

Principle #3Local quality

2Productivity

If only affine transformation is used for brain image normalization, then the processing is simple and fast, but the normalization accuracy of anatomical structures is insufficient

Engineering Contradiction:

Improvenormalization processing efficiencyVSAvoidanatomical landmark alignment precision

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent performs preliminary affine transformation to achieve basic alignment and normalize the overall brain shape before performing the more complex non-linear transformation on anatomical structures. This preliminary action reduces the complexity of the subsequent non-linear transformation and improves overall processing efficiency while maintaining high precision.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent segments the normalization process into two distinct stages: first performing affine transformation on the entire brain image to achieve basic shape normalization, then performing non-linear transformation on specific anatomical structures (eyes, diencephalon, fornix, corpus callosum, hippocampi) to achieve precise anatomical normalization. This segmentation allows each transformation method to be optimized for its specific target, resolving the contradiction between brain shape normalization and anatomical structure normalization precision.

Inventive Principle:

Principle #1Segmentation

3Manufacturing precision

If multiple transformation methods are applied to normalize both brain and anatomical structures, then the normalization accuracy is improved, but the processing complexity increases

Engineering Contradiction:

Improvenormalization accuracyVSAvoidnormalization process complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary affine transformation to achieve basic alignment and normalize the overall brain shape before performing the more complex non-linear transformation on anatomical structures. This preliminary action reduces the complexity of the subsequent non-linear transformation and improves overall processing efficiency while maintaining high precision.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent segments the normalization process into two distinct stages: first performing affine transformation on the entire brain image to achieve basic shape normalization, then performing non-linear transformation on specific anatomical structures (eyes, diencephalon, fornix, corpus callosum, hippocampi) to achieve precise anatomical normalization. This segmentation allows each transformation method to be optimized for its specific target, resolving the contradiction between brain shape normalization and anatomical structure normalization precision.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11257227B2Brain image normalization apparatus, brain image normalization method, and brain image normalization program

Publication Date: 2022.02.22 FUJIFILM CORP

AI summary

A brain image normalization apparatus, having a processor configured to: detect at least four reference landmarks of a left eye, a right eye, a diencephalon, a fornix, a corpus callosum, a left hippocampus, and a right hippocampus from a brain image including a brain of a subject; perform registration between the detected reference landmarks and reference landmarks corresponding to the detected reference landmarks included in a standard brain image; and normalize the brain image based on a result of the registration.