Tomosynthesis Reconstruction with Exponential Convolution Kernel
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional mammography methods produce 2D x-ray images that overlay tissue layers, making it difficult to detect tumors, while tomosynthesis provides limited 3D image quality due to incomplete scanning, resulting in reduced depth resolution and inability to quantify attenuation coefficients.
Innovation Solution
A tomosynthesis image reconstruction method using a discrete convolution kernel based on an exponential function for filtered back projection, which enhances image quality by emphasizing edge visualization and allowing diagnostic evaluation with minimal processing effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional 2D x-ray mammography is used, then the imaging process is simple and fast, but tissue layers are overlaid making tumor detection difficult
Solution Approach 1:
The patent transitions from 2D planar imaging to 3D volumetric imaging by introducing a tomographic dimension. Multiple projection images acquired at different angles are reconstructed into a 3D volume, allowing visualization of breast tissue in depth without overlay artifacts, thereby improving tumor detectability while maintaining imaging efficiency
2Loss of time
If tomosynthesis with restricted angular range is used, then processing time is reduced, but depth resolution deteriorates
Solution Approach 1:
The patent employs iterative reconstruction algorithms that optimize reconstruction parameters including angular sampling density, iteration count, and regularization strength. By adjusting these parameters, the system achieves acceptable depth resolution with fewer projection angles, thereby reducing processing time while maintaining diagnostic quality
Solution Approach 2:
The patent uses partial scanning with a restricted angular range (e.g., 15-30 degrees) rather than complete 360-degree scanning. This partial action approach provides sufficient depth resolution for diagnostic purposes while dramatically reducing acquisition and processing time compared to full tomographic scanning
3Measurement precision
If complete tomographic scanning is used, then depth resolution is improved, but processing effort and time increase significantly
Solution Approach 1:
The patent implements partial scanning with a limited angular range that provides sufficient depth resolution for clinical diagnosis without requiring complete tomographic coverage. This reduces the number of projection images needed, thereby decreasing processing complexity and computational burden while maintaining adequate depth resolution
Solution Approach 2:
The patent divides the breast volume into multiple thin slices or layers during reconstruction, processing each slice independently or with localized interpolation. This segmentation approach reduces the overall computational complexity by breaking down the large-scale 3D reconstruction problem into smaller, more manageable 2D reconstruction problems
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
The method effectively generates high-quality 3D x-ray images that improve tumor detection and visualization of microcalcifications and tumors, even with incomplete scanning data, by controlling contrast and noise behavior through the parameter 'a', and reduces image errors from truncated projections.
Implementation Method 1
a digital x-ray detector which records detector signals produced by an x-ray tube
Data Source
AI summary
In a tomosynthetic image reconstruction method and diagnostic device operating with such a method, a tomosynthetic 3D x-ray image is reconstructed by a discrete filtered back projection from a number of individual digital projection data recorded from different project angles within a restricted angular range, in which at least one filtering is performed with a convolution kernel that, in the local area outside of its central value, corresponds to an exponential function.