X-Ray Tomosynthesis With Neural-Net Enhancement For Thinner Slices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-resolution tomosynthesis machines face challenges in reducing acquisition and reconstruction times while maintaining image quality, as higher resolution detectors increase data collection and processing times, patient dose, and clinician review burdens.
Innovation Solution
A neural network-based system identifies a region of interest using machine learning to enhance resolution of selected data subsets, reducing unnecessary data collection and processing by employing a trained neural network to boost resolution of clinically relevant regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-resolution detectors are used to increase image resolution, then manufacturing precision is improved, but acquisition time increases and productivity deteriorates
Solution Approach 1:
The patent divides the image processing into two stages: first acquiring low-resolution data quickly across the entire field of view, then applying super-resolution algorithms to specific regions of interest. This segmentation allows the system to maintain high overall productivity while achieving high resolution where clinically necessary.
Solution Approach 2:
The patent applies different resolution qualities to different regions of the image. Full super-resolution is applied only to identified regions of interest rather than the entire image, optimizing the balance between manufacturing precision (resolution) and productivity (processing speed) by concentrating computational resources where they provide maximum diagnostic value.
2Manufacturing precision
If high-resolution detectors are used to increase image resolution, then manufacturing precision is improved, but data collection time increases and loss of time worsens
Solution Approach 1:
The patent performs preliminary acquisition of low-resolution data quickly, then applies super-resolution enhancement afterward. This preliminary action approach allows the system to collect data rapidly without the time penalty of high-resolution detection, and then enhance resolution computationally to achieve the desired manufacturing precision without the associated time cost.
3Manufacturing precision
If high-resolution detectors are used to increase image resolution, then manufacturing precision is improved, but processing time increases and productivity deteriorates
Solution Approach 1:
The patent extracts and processes only the necessary portions of the data at high resolution. By identifying regions of interest and applying super-resolution algorithms only to those specific areas rather than the entire dataset, the system maintains high manufacturing precision for clinically relevant regions while dramatically reducing overall processing time and improving productivity.
4Manufacturing precision
If high-resolution detectors are used to increase image resolution, then manufacturing precision is improved, but data volume increases and device complexity worsens
Solution Approach 1:
The patent applies partial super-resolution enhancement only to identified regions of interest rather than processing the entire dataset at maximum resolution. This partial action approach maintains the manufacturing precision needed for diagnostic accuracy in critical areas while reducing the overall data volume and computational complexity to manageable levels.
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 approach allows for faster acquisition and processing of high-resolution images, focusing on clinically relevant areas, thereby reducing unnecessary data and improving clinician efficiency.
Implementation Method 1
a trained neural network to accurately boost low-resolution images into high-resolution images after acquisition
Implementation Method 2
x-rays, for example, in a cone or parallel beam, is directed through a patient at a range of angles about the patient to be received by a multielement detector. At each angle, the detector collects 'projection attenuation data' representing the attenuation of x-ray photons
Data Source
Figure 1
Figure 2~3
Figure 4~6
AI summary
A tomosynthesis machine (10) allows for faster image acquisition and improved signal-to-noise by acquiring a projection attenuation data (61) and using machine learning (86) to identify a subset of the projection attenuation data (70) for the production of thinner slices and/or higher resolution slices (76) using machine learning.