Breast MRI Artifact Region Segmentation via Time-Variability Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for processing temporally acquired image data, such as in breast MRI scans, face challenges in accurately distinguishing between tumor structures and artifacts like blood vessels, leading to potential false identifications due to movement and contrast agent effects.
Innovation Solution
A method that computes a time-variability map from the image data, classifies locations based on this map, and uses projection and filtering techniques to separate artifact and non-artifact regions, thereby reducing the risk of misidentifying artifacts as tumor structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If variance processing is used to enhance lesion detection, then tumor detection capability is improved, but false identification of blood vessels as tumors increases
Solution Approach 1:
The patent segments the image processing by creating separate artifact regions and non-artifact regions based on time-variability characteristics. By dividing the processing space into these distinct regions, the system can apply different handling strategies: artifact regions are excluded from tumor detection while non-artifact regions undergo standard variance processing for lesion enhancement, thereby resolving the contradiction between detection capability and false identification.
Solution Approach 2:
The patent applies local quality by assigning different properties to different spatial regions. Artifact regions are identified through time-variability analysis and assigned the property of being excluded from tumor detection, while non-artifact regions retain the property of undergoing variance processing. This localized differentiation allows the system to maintain high tumor detection accuracy in relevant areas while suppressing false positives in artifact-prone areas.
2Productivity
If computer automated detection is used to identify lesions, then processing efficiency is improved, but accuracy decreases due to false identification of artifacts
Solution Approach 1:
The patent implements preliminary action by performing artifact region identification and exclusion before the main tumor detection process. The system first computes time-variability maps to identify artifact locations, creates artifact region masks, and then applies these masks to exclude artifact regions from subsequent variance processing and tumor detection. This preliminary separation ensures that the automated detection system processes only relevant non-artifact regions, maintaining both efficiency and accuracy.
3Area of stationary object
If the thoracic cage region is included in breast MRI scans, then comprehensive anatomical coverage is improved, but artifact generation from heart and aorta movement increases
Solution Approach 1:
The patent applies the taking out principle by extracting and isolating artifact regions from the comprehensive anatomical coverage. The system identifies regions containing moving structures like heart and aorta through time-variability analysis, separates these artifact-prone regions from the rest of the anatomical coverage, and excludes them from tumor detection processing. This allows the system to maintain comprehensive anatomical coverage for reference while preventing artifact generation from interfering with detection accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method (100) of processing temporally acquired image data comprising an obtaining step (105) for obtaining the temporally acquired image data, a computing step (110) for computing a time- variability map on the basis of the temporally acquired image data, a classifying step (120) for classifying locations of the temporally acquired image data on the basis of the time-variability map, and a determining step (125) for determining an artifact region and a non-artifact region in the temporally acquired image data on the basis of the classified locations. After determining the artifact region and the non-artifact region, detecting an object in a detecting step (130) is limited to the non-artifact region. This advantageously reduces the risk of falsely identifying the detected object as an object of interest.