3D Tomography Reconstruction Pipeline for Low-Dose X-Ray Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Tomographic images generated from x-ray machines are noisy, which reduces their diagnostic value, and increasing the x-ray radiation dose to reduce noise is detrimental to the subject being scanned.
Innovation Solution
A 3D tomography reconstruction pipeline that includes a first neural network model, a fixed function backprojection unit, and a second neural network model processes tomography images to produce a reconstructed 3D density volume, reducing noise and allowing for lower x-ray doses by integrating noise reduction techniques such as pre-filtering and self-supervised training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the x-ray radiation dose is increased to reduce noise, then the image quality is improved, but the harm to the subject is increased
Solution Approach 1:
The patent replaces the traditional mechanical approach of increasing radiation dose to improve image quality with a computational approach using neural networks. The system uses deep learning models to denoise low-dose CT images, substituting the physical mechanism of high-dose radiation with an information-processing mechanism that analyzes and reconstructs image data algorithmically.
Solution Approach 2:
The patent introduces neural network models as intermediaries between the raw low-dose CT images and the final diagnostic images. These neural networks act as mediators that process the noisy input images and produce high-quality output images without requiring high radiation doses, thus resolving the contradiction between image quality and radiation harm.
2Measurement precision
If traditional reconstruction methods are used, then the processing is simple, but the noise reduction capability is insufficient
Solution Approach 1:
The patent segments the image reconstruction process into multiple specialized neural network components: a first neural network for initial denoising, a backprojection unit for 3D reconstruction, and a second neural network for final refinement. This segmentation allows each component to focus on specific aspects of noise reduction while maintaining overall system manageability through modular architecture.
Solution Approach 2:
The patent transitions from 2D slice-by-slice reconstruction to 3D volumetric reconstruction by introducing a backprojection unit that operates in the third dimension. This dimensional change enables the system to leverage spatial correlations across multiple slices for more effective noise reduction, overcoming the limitations of traditional 2D methods.
Data Source
AI summary
A three-dimensional (3D) density volume of an object is constructed from tomography images (e.g., x-ray images) of the object. The tomography images are projection images that capture all structures of an object (e.g., human body) between a beam source and imaging sensor. The beam effectively integrates along a path through the object producing a tomography image at the imaging sensor, where each pixel represents attenuation. A 3D reconstruction pipeline includes a first neural network model, a fixed function backprojection unit, and a second neural network model. Given information for the capture environment, the tomography images are processed by the reconstruction pipeline to produce a reconstructed 3D density volume of the object. In contrast with a set of 2D slices, the entire 3D density volume is reconstructed, so two-dimensional (2D) density images may be produced by slicing through any portion of the 3D density volume at any angle.