Brain Tissue Classification via Multi-Volume MRI Fusion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for automatically detecting and delineating boundaries between brain tissue and cerebrospinal fluid (CSF) regions in medical images, particularly MRI, face challenges such as noise, limited resolution, limited contrast, inhomogeneities, and difficulties in distinguishing thin sulci and brain pathologies like edema, which complicate accurate tissue classification.

Innovation Solution

The method involves combining data from two or more MRI volumes with varying acquisition parameters to enhance contrast, using thresholding and iterative normalization to correct inhomogeneities, and employing a skeletonization algorithm to detect thin sulcal boundaries, while leveraging multiple imaging modalities for improved accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If single MRI volume is used for tissue classification, then processing is simple and fast, but contrast between tissue types is limited and inhomogeneities are present

Engineering Contradiction:
Improvetissue classification accuracyVSAvoidimage processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines multiple MRI volumes acquired with different acquisition parameters (e.g., different flip angles or echo times) to create an enhanced classification image. By merging information from multiple sources, the method achieves improved tissue contrast and reduced inhomogeneities while maintaining automated processing, thus resolving the contradiction between measurement precision and device complexity.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The method utilizes multiple MRI volumes with varying acquisition parameters (such as flip angle, echo time, or repetition time) to obtain different tissue contrast characteristics. By changing these parameters across multiple scans and combining the results, the patent enhances tissue differentiation without requiring complex manual intervention, addressing both accuracy improvement and processing simplicity.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If thresholding method is used for tissue classification, then processing is fast and simple, but boundaries are not accurately located especially for thin sulci

Engineering Contradiction:
Improveboundary detection accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary image enhancement by combining multiple MRI volumes with different acquisition parameters before applying thresholding. This preliminary action improves the contrast and sharpness of tissue boundaries, enabling accurate detection of thin sulci and precise boundary location while maintaining the speed and simplicity of threshold-based classification.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The method transforms the classification problem by working in an enhanced image domain created from multiple volumes. By adding the dimension of multiple acquisition parameters and combining them through weighted averaging or other fusion techniques, the patent achieves superior boundary detection accuracy without significantly increasing processing time, as the enhancement is performed once before thresholding.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Measurement precision

If deformable surface models are used to represent tissue boundaries, then partial voxels can be represented, but implementation is complicated and computation is intensive

Engineering Contradiction:
Improveboundary representation accuracyVSAvoidalgorithm complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts and utilizes only the essential information from multiple MRI volumes (enhanced contrast and boundary sharpness) to improve thresholding effectiveness. By taking out only the necessary features from the multi-volume data and applying them to enhance the thresholding process, the method achieves accurate boundary representation without implementing complex deformable surface models, thus reducing algorithmic complexity while maintaining precision.

Inventive Principle:
Principle #2Taking out (Extraction)

4Measurement precision

If intensity normalization is applied to correct inhomogeneities, then tissue contrast is improved, but processing steps increase and time is consumed

Engineering Contradiction:
Improvetissue contrastVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs intensity normalization as a preliminary action when combining multiple MRI volumes. By normalizing the intensities of individual volumes before fusion and incorporating normalization into the combination process itself, the method achieves improved tissue contrast while minimizing additional processing time, as the normalization is integrated rather than applied as a separate post-processing step.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP1768041B1Brain tissue classification
Publication Date: 2016.04.13 BRAINLAB AG
  • EP1768041B1 patent drawingFigure 1
  • EP1768041B1 patent drawingFigure 2
  • EP1768041B1 patent drawingFigure 3

AI summary

The application relates to a method of medical image processing to automatically classify the major tissue types, white matter (WM), gray matter (GM), and cerebro-spinal fluid (CSF), particularly using images generated by magnetic resonance imaging (MRI). Noise, limited resolution, limited contrast between tissue types, inhomogeneities in the measurement device, and tissue variation can complicate this problem. The invention improves contrast and reduces inhomogeneities by combining data from two or more similar data sets, varying at least one acquisition parameter to obtain different tissue contrast in the different volumes. The methods for choosing the acquisition parameters and for combining the images are elements of the invention. The classification method is further optimized by adjusting the volume signal levels so that the classification boundaries tend to lie on edges in the original image.