X-ray Detector Noise Reduction via Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital x-ray imaging systems face challenges with row-correlated noise artifacts caused by electromagnetic interference, which traditional shielding methods are ineffective in addressing, leading to reduced image quality and increased system complexity.
Innovation Solution
A neural network model is trained using pairs of noisy and clean images to identify and remove electromagnetic interference noise, with the image being subdivided into segments to enhance training efficiency and noise reduction, allowing for real-time correction of acquired images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If traditional electromagnetic shielding methods are used, then electromagnetic interference is reduced, but system complexity and cost increase
Solution Approach 1:
The patent replaces physical electromagnetic shielding mechanisms with a computational neural network model. Instead of using bulky metallic shields or complex electromagnetic interference suppression hardware, the system uses a trained neural network to identify and remove noise artifacts from images, substituting a mechanical/physical solution with an information-processing solution.
Solution Approach 2:
The neural network model acts as an intermediary between the noisy image data and the final clean image output. It processes the image data, learns to distinguish between actual anatomical structures and noise artifacts, and produces corrected images without requiring complex physical shielding infrastructure.
2Manufacturing precision
If neural network model is used for noise reduction, then image quality is improved, but processing time increases
Solution Approach 1:
The neural network model is trained in advance using pairs of noisy and clean images. During training, the model learns to recognize patterns and relationships between noise artifacts and corresponding anatomical structures. Once trained, the model can quickly process new images without requiring time-consuming training, achieving fast inference while maintaining high image quality.
3Productivity
If image is subdivided into segments, then training efficiency is enhanced, but image complexity increases
Solution Approach 1:
The patent divides the image into multiple segments or patches that can be processed independently by the neural network. This segmentation allows the model to focus on smaller regions at a time, improving training efficiency and enabling parallel processing. The segments are then reassembled to form the complete corrected image, maintaining overall image integrity while simplifying the processing complexity.
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 solution effectively reduces row-correlated noise artifacts, improving image quality without the need for extensive electromagnetic shielding, and enables faster convergence of the neural network model, resulting in enhanced performance and image clarity.
Implementation Method 1
A neural network model is trained using pairs of noisy and clean images to identify and remove electromagnetic interference noise
Implementation Method 2
A portion of the detector converts the radiation to light photons that are sensed
Data Source
AI summary
Various methods and systems are provided for x-ray imaging. In one embodiment, a method for an x-ray imaging system comprises acquiring, with an x-ray detector, an image including a noise artifact caused by electromagnetic interference, inputting the image to a trained neural network model to obtain a corrected image with the noise artifact removed, and outputting the corrected image. In this way, row-correlated noise artifacts caused by electromagnetic interference at the x-ray detector are eliminated or cancelled in real time and image quality is improved.