Magnetic Resonance Image Mask Generation Using Magnitude Phase Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current image segmentation methods for magnetic resonance imaging, particularly in cerebral cortex segmentation, face challenges with low gray/white matter contrast in older and diseased populations, leading to poor image quality and noise pollution.

Innovation Solution

A method that combines magnitude and phase information from T2*-weighted data using K-means clustering and an iterative framework to obtain image masks, followed by homodyne high-pass filtering and phase unwrapping, and binarization threshold processing to accurately distinguish foreground and background pixels.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If binary threshold method (Otsu's method) is used for image segmentation, then the segmentation process is simple and fast, but the segmentation accuracy deteriorates in regions with low gray/white matter contrast

Engineering Contradiction:
Improvesegmentation processing speedVSAvoidsegmentation accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent combines magnitude image information with phase image information to create a composite image for segmentation. This merging of multiple information sources compensates for the low contrast problem in magnitude images alone, improving segmentation accuracy without significantly increasing processing complexity

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent transforms the phase image through homodyne high-pass filtering and phase unwrapping operations to enhance the contrast and visibility of gray/white matter boundaries. These parameter transformations improve the effectiveness of threshold-based segmentation in low-contrast regions

Inventive Principle:
Principle #35Parameter changes

2Device complexity

If conventional magnitude image-based segmentation is used, then the processing framework is simple, but the segmentation quality deteriorates due to noise pollution and low contrast in older and diseased populations

Engineering Contradiction:
Improveprocessing framework complexityVSAvoidsegmentation quality
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent integrates magnitude and phase information into a unified segmentation framework. This combination provides complementary information that enhances reliability by compensating for the limitations of magnitude images alone, particularly in pathological cases with low tissue contrast

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent introduces phase information as an intermediary element that mediates the segmentation process. The phase data serves as additional evidence to disambiguate regions where magnitude information is insufficient, improving overall segmentation reliability

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3147863B1Obtaining image mask
Publication Date: 2020.09.23 SHANGHAI NEUSOFT MEDICAL TECH LTD
  • EP3147863B1 patent drawingFigure 1
  • EP3147863B1 patent drawingFigure 2A~2B
  • EP3147863B1 patent drawingFigure 2C~2D

AI summary

In an example, a method and apparatus for obtaining an image mask is provided. After a magnitude image and a phase image of a to-be-processed image is obtained, magnitude coherent data of each pixel point in the magnitude image and phase coherent data of each pixel point in the phase image may be calculated. Then, a binarization threshold processing may be performed on the magnitude coherent data of each pixel point in the magnitude image to obtain a magnitude image mask. A binarization threshold processing may be performed on the phase coherent data of each pixel point in the phase image to obtain a phase image mask. In this way, an image mask of the to-be-processed image may be obtained by using the magnitude image mask and the phase image mask.