Multi-Stage Denoising Algorithm for Cone Beam CT Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional denoising techniques for CT imaging, such as digital filter-based and wavelet-transform-based methods, are inefficient in reducing noise while preserving image spatial resolution and contrast, particularly in clinical data, and increase computational time, making them impractical for reducing radiation exposure dose effectively.
Innovation Solution
A multi-stage denoising algorithm combining digital reconstruction filter (DRF) and wavelet-transform (WT) techniques in the projection domain of cone beam CT imaging, using Fourier transforms and adaptive noise reduction stages to optimize noise reduction without compromising image quality, incorporating feedback mechanisms for optimal performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional denoising techniques (digital filter-based or wavelet-transform-based) are applied to final reconstruction images, then noise level is reduced, but image spatial resolution and contrast are degraded and computational time increases
Solution Approach 1:
The patent applies denoising filters to projection data before image reconstruction rather than to final reconstructed images. This preliminary denoising of projection data prevents noise from being amplified during the reconstruction process, thereby reducing noise in the final image while preserving spatial resolution and contrast that would otherwise be degraded by post-reconstruction denoising.
Solution Approach 2:
The patent implements a multi-stage denoising process that segments the denoising operation into distinct stages: first stage applies a digital reconstruction filter (DRF) to projection data, and a second stage applies wavelet-transform-based denoising to the DRF-filtered projection data. This segmentation allows each stage to address different aspects of noise while preserving image quality.
2Object-affected harmful factors
If strong denoising filters are applied to reduce noise significantly, then noise level is reduced, but high frequency information is filtered out resulting in reduced contrast and sharpness
Solution Approach 1:
The patent employs different denoising filters with different characteristics in a multi-stage process: the digital reconstruction filter (DRF) provides initial noise reduction while the wavelet-transform-based filter provides additional denoising. Each filter operates with appropriate strength for its stage, preventing excessive filtering that would remove high frequency information while still achieving significant overall noise reduction.
Solution Approach 2:
The patent combines multiple denoising approaches (digital reconstruction filtering and wavelet-transform-based filtering) into a composite denoising system. This composite approach leverages the strengths of each method to achieve superior noise reduction while preserving high frequency information that would be lost using either method alone with strong filtering.
3Object-affected harmful factors
If multi-stage denoising with both DRF and WT is applied to projection data, then noise is reduced while preserving image quality, but device complexity increases
Solution Approach 1:
The patent divides the denoising process into two distinct stages applied to projection data: first stage uses digital reconstruction filtering and the second stage uses wavelet-transform-based filtering. This segmentation allows each stage to be optimized independently and provides clarity in the processing flow, making the overall complex system more manageable and implementable.
4Object-affected harmful factors
If conventional denoising is applied to final reconstruction images, then noise is reduced, but reconstruction time increases making it impractical for clinical use
Solution Approach 1:
The patent performs denoising on projection data before the computationally intensive image reconstruction process. By filtering projection data preliminarily, the denoising operations are more efficient and prevent noise from propagating through the reconstruction algorithm, thereby reducing overall computational time compared to applying denoising to final reconstructed images.
Solution Approach 2:
The patent implements a two-stage denoising process on projection data: first stage applies digital reconstruction filtering and second stage applies wavelet-transform-based filtering. This segmented approach processes data in manageable stages before reconstruction, improving computational efficiency and reducing total reconstruction time compared to post-reconstruction denoising.
Data Source
AI summary
Raw cone beam tomography projection image data are taken from an object and are denoised by a wavelet domain denoising technique and at least one other denoising technique such as a digital reconstruction filter. The denoised projection image data are then reconstructed into the final tomography image using a cone beam reconstruction algorithm, such as Feldkamp's algorithm.