Tomosynthetic Image Reconstruction for Microcalcification Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional mammography methods face challenges in detecting micro-calcifications due to high image noise in 2D projection images, which can lead to misdiagnosis when noise-reducing techniques obscure these diagnostic features in tomosynthetic 3D x-ray images.
Innovation Solution
A tomosynthetic image reconstruction method that generates a 3D x-ray image by reconstructing an intermediate image without noise filtering, segmenting microcalcium regions, selecting relevant slice images, forward projecting microcalcium voxels, and applying reduced noise filtering only to these pixels, allowing for improved detection of microcalcifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If noise-reducing image processing is applied to 2D projection images before reconstruction, then image noise is reduced, but micro-calcifications disappear or merge together leading to misdiagnosis
Solution Approach 1:
The patent segments the image processing into two distinct pathways: one for micro-calcification regions and another for the remaining image regions. Micro-calcification pixels are identified through clustering algorithms and processed separately with reduced or no noise filtering, while other regions receive standard noise reduction. This segmentation allows preserving diagnostic features while reducing noise in non-critical areas.
Solution Approach 2:
The patent applies different noise filtering strengths to different spatial regions of the image. Specifically, micro-calcification pixels receive minimal or no noise filtering (local quality preservation), while non-microcalcification regions receive stronger noise reduction (local quality improvement). This localized differential processing resolves the contradiction by tailoring the noise reduction intensity to the diagnostic importance of each region.
2Use of energy by moving object
If individual 2D projection images are acquired with low dose to stay within total dose limits, then total radiation dose is controlled, but image noise increases making reconstruction difficult
Solution Approach 1:
The patent performs preliminary identification and marking of micro-calcification pixels in the low-dose projection images before reconstruction. By pre-identifying these critical features through clustering analysis and forward projection, the system prepares a map of regions that require special handling during noise reduction, enabling subsequent selective preservation of these features even in low-dose conditions.
Solution Approach 2:
The patent changes the noise filtering parameter (filtering strength) dynamically based on the pixel type. For micro-calcification pixels, the filtering parameter is set to minimal or zero strength, while for other pixels, stronger filtering is applied. This parameter adaptation allows the system to maintain low radiation dose while preserving critical diagnostic features through selective parameter adjustment.
3Object-affected harmful factors
If micro-calcification pixels are subjected to strong noise filtering, then noise is reduced in those regions, but micro-calcifications may be obscured or merged
Solution Approach 1:
The patent applies preliminary anti-action by identifying micro-calcification pixels before noise filtering and marking them for protected processing. This pre-identification creates a protective mechanism that prevents these critical pixels from being subjected to strong noise filtering that would obscure them. The forward projection and clustering algorithms establish this protective map in advance, counteracting the potential harmful effect of uniform noise filtering.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method enables the generation of low-noise 3D x-ray images that effectively identify diagnostically relevant microstructures, enhancing the detection of microcalcifications in specific slices while minimizing noise impact.
Implementation Method 1
In conventional x-ray mammography, a two-dimensional single image of the compressed breast is generated in a single projection direction.
Implementation Method 2
a tomosynthetic 3D x-ray image is assembled from a number of digital 2D projection images acquired from various projection angles
Data Source
AI summary
To generate a tomosynthetic 3D x-ray image composed of a number of slice images, a tomosynthetic 3D intermediate image composed of a number of slice images is reconstructed from 2D projection images that have not been noise-filtered. The microcalcium regions contained in these slice images are segmented, and one or more subject slice images relevant to these microcalcium voxels are determined for each microcalcium region. The microcalcium voxels belonging to the segmented microcalcium region in this subject slice image or in these subject slice images are projected forwards in the 2D projection images, and the microcalcium pixels associated with these microcalcium voxels are marked in the 2D projection images. Noise-filtered 2D projection images are subsequently generated by subjecting the microcalcium pixels of the 2D projection images to no noise filtering or a noise filtering that leads to a noise reduction reduced relative to the remaining image regions. The tomosynthetic 3D x-ray image is then calculated from the 2D projection images that have been noise-filtered in this manner.