Noise Preserving Sharpening Filter for CT Resolution Recovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for sharpening medical CT images often amplify high-frequency noise, leading to artifacts and a lack of clinically important detail, while attempts to denoise and sharpen simultaneously can result in overly smooth images.
Innovation Solution
A system and method that utilize a deep neural network-based noise preserving sharpening filter (NPSF) to sharpen CT images while preserving noise energy, using adjustable parameters to achieve a suitable noise-resolution tradeoff.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sharpening methods are applied to CT images, then image resolution is improved, but noise is substantially amplified leading to artifacts
Solution Approach 1:
The patent converts the harmful noise amplification effect into a beneficial feature by training the neural network to preserve noise characteristics. The network learns to distinguish between noise and anatomical structures, allowing resolution enhancement while maintaining original noise levels rather than amplifying them.
Solution Approach 2:
The patent changes the optimization parameters of the neural network by using a composite loss function that includes both reconstruction error and noise preservation terms. This parameter change allows the network to simultaneously improve resolution and control noise amplification through weighted optimization.
2Object-generated harmful factors
If denoising is applied before sharpening, then noise is suppressed, but clinically important detail is lost
Solution Approach 1:
The patent performs sharpening and noise preservation simultaneously in a single neural network pass rather than sequentially applying denoising then sharpening. This preliminary integrated action prevents the loss of clinically important detail that occurs when denoising is applied first.
Solution Approach 2:
The patent applies different processing characteristics to different regions of the image by using local noise variance estimation. The network adapts its sharpening and noise preservation behavior locally based on the specific noise characteristics and anatomical content of each region.
3Measurement precision
If combined denoising and sharpening is applied, then noise is reduced and resolution is improved, but images become overly smooth
Solution Approach 1:
The patent introduces adjustable parameters that allow dynamic control of the sharpening and noise preservation balance. Users can adjust the sharpening strength and noise preservation weight to achieve the desired level of detail preservation without excessive smoothness.
Solution Approach 2:
The patent incorporates feedback mechanisms through the loss function that monitors both reconstruction accuracy and noise preservation. This feedback guides the network to maintain appropriate texture characteristics while enhancing resolution, preventing overly smooth results.
Data Source
AI summary
Noise preserving models and methods for resolution recovery of x-ray computed tomography (e.g., using a computerized tool) are enabled. For example, a system can comprise: a memory that stores computer executable components, and a processor that executes the computer executable components stored in the memory, wherein the computer executable components comprise: a pair generation component that generates a pair of images, the pair of images comprising an input image and a ground truth image, a training component that trains a machine learning based sharpening algorithm by approximately minimizing a loss function that determines an error between a sharpened image and the ground truth image, and a sharpening component that, using the sharpening algorithm, sharpens the input image to generate the sharpened image, wherein the sharpened image comprises a second noise that is similar in intensity to a first noise of the input image.