Tomosynthesis CAD Using 2D-to-3D Back-Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Tomosynthesis mammography generates a large amount of data, making it challenging for radiologists to interpret effectively, as traditional CAD algorithms only process either 2-D or 3-D images separately, lacking a comprehensive approach to visualize and analyze regions of interest across both dimensions.
Innovation Solution
Combining 2-D and 3-D image processing methods to detect and analyze regions of interest in tomosynthesis mammography data, using back-projection to integrate findings from 2-D projection images into 3-D volume data, and employing multiplanar reformatting and volume rendering for synchronized visualization across multiple data sets and views.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If tomosynthesis mammography is used to generate 3-D images, then diagnostic information and detection accuracy are improved, but the amount of data increases making interpretation more difficult
Solution Approach 1:
The patent segments the large 3-D tomosynthesis data set into multiple 2-D projection images at different angles. This segmentation allows radiologists to analyze the data in manageable 2-D slices rather than overwhelming 3-D volume, while still maintaining the diagnostic benefits of tomosynthesis through the ability to reconstruct and view multiple planar reformats.
Solution Approach 2:
The patent transforms the 3-D volume data into multiple 2-D projection images, effectively changing the dimensionality representation. This allows the system to present 3-D diagnostic information in a 2-D format that is more familiar and easier to interpret for radiologists, resolving the contradiction between improved diagnostic accuracy and manageable data presentation.
2Device complexity
If traditional CAD algorithms process only 2-D or 3-D images separately, then processing complexity is reduced, but comprehensive analysis of regions of interest is limited
Solution Approach 1:
The patent merges 2-D and 3-D processing approaches by using 2-D projection images to identify candidate regions of interest, then back-projecting these candidates into the 3-D volume for comprehensive analysis. This combination allows the system to leverage the simplicity of 2-D processing while achieving the comprehensive diagnostic capability of 3-D analysis, resolving the contradiction between processing complexity and information completeness.
Solution Approach 2:
The patent introduces back-projection as an intermediary technique that connects 2-D projection data with 3-D volume data. By using back-projection to map 2-D detected regions into the 3-D space, the system enables comprehensive 3-D analysis without requiring direct complex 3-D processing of the entire data set, thus maintaining manageable complexity while achieving complete analysis.
Data Source
AI summary
The present invention provides a method and system using computer-aided detection (CAD) algorithms to aid diagnosis and visualization of tomosynthesis mammography data. The proposed CAD algorithms process two-dimensional and three-dimensional tomosynthesis mammography images and identify regions of interest in breasts. The CAD algorithms include the steps of preprocessing; candidate detection of potential regions of interest; and classification of each region of interest to aid reading by radiologists. The detection of potential regions of interest utilizes two dimensional projection images for generating candidates. The resultant candidates in two dimensional images are back-projected into the three dimensional volume images. The feature extraction for classification operates in the three dimensional image in the neighborhood of the back-projected candidate location. The forward-projection and back-projection algorithms are used for visualization of the tomosynthesis mammography data in a fashion of synchronized MPR and VR.